Prediction device, communication control device, and prediction method

The prediction device classifies communication traffic and uses tailored prediction methods to estimate future traffic volume, addressing the challenges of computational load and delay in conventional systems, thereby reducing congestion and communication delays.

WO2024236755A1PCT designated stage expired Publication Date: 2024-11-21NIPPON TELEGRAPH & TELEPHONE CORP
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
PCT/JP2023/018394
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-05-17
Publication Date
2024-11-21

AI Technical Summary

Technical Problem

Conventional communication systems face challenges in predicting future traffic volume accurately across multiple flows, leading to increased computational load and delayed route switching, which can result in communication delays and congestion.

Method used

A prediction device and method that classify communication traffic into flows based on characteristics, using a combination of low and high calculation prediction methods to estimate future traffic volume, reducing the overall computational burden and enabling timely route switching.

Benefits of technology

This approach effectively suppresses congestion and reduces communication delays while maintaining communication quality, even in environments with multiple coexisting flows.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure JP2023018394_21112024_PF_FP_ABST
    Figure JP2023018394_21112024_PF_FP_ABST
Patent Text Reader

Abstract

This prediction device comprises: a flow classification unit that classifies communication traffic, which is input during a predetermined period, into a plurality of flows on the basis of characteristics of the communication traffic; a first prediction unit that predicts future traffic volumes of the flows by using a first prediction method on the basis of current traffic volumes of the flows; a second prediction unit that predicts future traffic volumes of the flows by using a second prediction method that requires less calculations than the first prediction method on the basis of the current traffic volumes of the flows; a prediction method determination unit that determines a prediction method to be used for each of the flows on the basis of the characteristics of the communication traffic; and a total traffic calculation unit that calculates the total value of the traffic volume for each of the flows predicted by the first prediction unit and the traffic volume for each of the flows predicted by the second prediction unit.
Need to check novelty before this filing date? Find Prior Art

Description

Prediction device, communication control device, and prediction method

[0001] The present invention relates to a prediction device, a communication control device, and a prediction method.

[0002] When multiple communication devices communicate simultaneously, such as when multiple upper devices and multiple lower devices communicate with each other, congestion may occur, resulting in communication delays. Congestion may be resolved by switching the communication path (hereinafter referred to as "path switching") using a switching device installed on the communication path, thereby changing the connection relationship between the upper device and the lower device. If communication traffic (hereinafter simply referred to as "traffic") concentrated on a specific communication device (e.g., an upper device) can be distributed to multiple communication devices by path switching, congestion can be resolved, and communication delays can be reduced.

[0003] Conventionally, in a communication system having a switching device that can change the connection relationship between a higher-level device and a lower-level device, there is a technology that collects information indicating the amount of traffic measured by the higher-level device and the lower-level device (hereinafter simply referred to as "traffic volume"), determines whether congestion has occurred at the offload destination based on the collected traffic volume, and performs route switching.

[0004] JP2014-060610A JP10-285276A

[0005] Y.Okamoto, et al., “Short-term traffic prediction based on mobile control information for proactive optical switching to lower congestion delay,” International Conference on Emerging Technologies for Communications (ICETC), 2022

[0006] Conventional communication systems calculate a congestion value indicating the degree of congestion based on the traffic volume of each communication path, and perform route switching when the congestion value of any communication path exceeds a threshold. However, route switching generally requires a certain amount of time. Therefore, in conventional communication systems, congestion occurs from the time congestion occurs until the route switching is completed. Therefore, the time required for route switching is also one of the factors that cause communication delays. For example, in the technologies described in Patent Documents 1 and 2, it takes time from collecting actual traffic volume to performing route switching, making it difficult to determine whether congestion has occurred in real time (e.g., within several milliseconds) and to perform route switching within a predetermined period of time (e.g., within several tens of milliseconds).

[0007] In response to this, a method can be considered in which, instead of switching routes based on actual measured values ​​of traffic volume, future traffic volume is predicted and route switching is performed in advance based on the predicted value to avoid congestion. However, in an actual communication environment in which multiple traffic flows (hereinafter simply referred to as "flows") with various traffic characteristics coexist, as in the case of communications with various QoS (Quality of Service) being performed between a large number of communication devices these days, there is a problem in that predicting the future traffic volumes of all flows collectively results in low prediction accuracy.

[0008] On the other hand, when future traffic volume is predicted for each flow, the computational load associated with the prediction increases in proportion to the number of flows. This increases the computation time required for the prediction, leading to the problem that route switching cannot be completed before congestion actually occurs. Thus, in the prior art, even if a decision to perform route switching is made based on predicted traffic volume, congestion cannot be prevented, resulting in communication delays.

[0009] In view of the above circumstances, the present invention aims to provide a prediction device, a communication control device, and a prediction method that can reduce communication delay while satisfying communication quality even when multiple flows are mixed.

[0010] One aspect of the present invention is a prediction device comprising: a flow classification unit that classifies communication traffic input during a predetermined period into a plurality of flows based on characteristics of the communication traffic; a first prediction unit that predicts a future traffic volume of the flow using a first prediction method based on a current traffic volume of the flow; a second prediction unit that predicts a future traffic volume of the flow using a second prediction method that requires less computational effort than the first prediction method based on the current traffic volume of the flow; a prediction method determination unit that determines a prediction method to be used for each of the flows based on the characteristics of the communication traffic; and a total traffic calculation unit that calculates the sum of the traffic volume for each flow predicted by the first prediction unit and the traffic volume for each flow predicted by the second prediction unit.

[0011] Another aspect of the present invention is a communication control device comprising: a flow classification unit that classifies communication traffic input during a predetermined period into a plurality of flows based on characteristics of the communication traffic; a first prediction unit that predicts a future traffic volume of the flow using a first prediction method based on a current traffic volume of the flow; a second prediction unit that predicts a future traffic volume of the flow using a second prediction method that requires less calculations than the first prediction method based on the current traffic volume of the flow; a prediction method determination unit that determines a prediction method to be used for each of the flows based on the characteristics of the communication traffic; a total traffic calculation unit that calculates the sum of the traffic volume for each flow predicted by the first prediction unit and the traffic volume for each flow predicted by the second prediction unit; a switching determination unit that predicts whether congestion will occur in the future based on the sum and determines whether a communication path needs to be switched based on the prediction result; and a switching instruction unit that, when the switching determination unit determines that a communication path needs to be switched, issues a switching instruction to a switching device that performs the communication path switching.

[0012] Another aspect of the present invention is a prediction method executed by a computer, the prediction method including: a flow classification step of classifying communication traffic input during a predetermined period into a plurality of flows based on characteristics of the communication traffic; a first prediction step of predicting a future traffic volume of the flow using a first prediction method based on a current traffic volume of the flow; a second prediction step of predicting a future traffic volume of the flow using a second prediction method that requires less computational effort than the first prediction method based on the current traffic volume of the flow; a prediction method determination step of determining a prediction method to be used for each of the flows based on the characteristics of the communication traffic; and a total traffic calculation step of calculating the sum of the traffic volume for each flow predicted in the first prediction step and the traffic volume for each flow predicted in the second prediction step.

[0013] According to the present invention, even when a plurality of flows are mixed, it is possible to suppress the occurrence of congestion and reduce communication delays while satisfying communication quality.

[0014] 1 is a diagram showing the configuration of a communication system 1, which is an example of a conventional communication system. FIG. 2 is a diagram showing the configuration of a mobile communication system 10, which is an example of a conventional communication system. FIG. 3 is a block diagram showing the configuration of a switching instruction device 17 of the mobile communication system 10, which is an example of a conventional communication system. FIG. 4 is a diagram showing the flow of processing in the switching instruction device 17 of the mobile communication system 10, which is an example of a conventional communication system. FIG. 5 is a flowchart showing the processing for predicting traffic volume of a remote station in a prediction unit 171. FIG. 6 is a flowchart showing the processing for creating a traffic volume prediction model in the prediction unit 171. FIG. 7 is a flowchart showing the processing for predicting future traffic volume in the prediction unit 171. FIG. 8 is a flowchart showing the processing for predicting traffic volume of an upper link in the prediction unit 171. FIG. 9 is a flowchart showing the processing for calculating a switching threshold in a switching threshold determination unit 172. FIG. 10 is a diagram showing an example of a lookup table. FIG. 11 is a flowchart showing the switching judgment processing in a switching judgment unit 175. FIG. 11 is a block diagram showing the functional configuration of a prediction unit 171a of a communication system according to a first embodiment of the present invention. FIG. 12 is a schematic diagram for explaining the definition of a flow, which is a prerequisite in the first embodiment of the present invention. FIG. 13 is a diagram showing an example of the configuration of a flow information table in the first embodiment of the present invention. FIG. 14 is a block diagram showing the functional configuration of a prediction execution unit 730 of a communication system according to the first embodiment of the present invention. FIG. 10 is a block diagram showing the functional configuration of a prediction unit 171b of a communication system according to a second embodiment of the present invention. FIG. 11 is a flowchart showing the operation of a prediction execution unit 730b of a communication system according to the second embodiment of the present invention. FIG. 12 is a block diagram showing the functional configuration of a prediction unit 171c of a communication system according to a third embodiment of the present invention. FIG. 13 is a flowchart showing the operation of a prediction execution unit 730c of a communication system according to the third embodiment of the present invention. FIG. 14 is a block diagram showing the functional configuration of a prediction unit 171d of a communication system according to a fourth embodiment of the present invention. FIG. 15 is a flowchart showing the operation of a prediction execution unit 730d of a communication system according to the fourth embodiment of the present invention. FIG. 16 is a block diagram showing the functional configuration of a prediction unit 171e of a communication system according to a fifth embodiment of the present invention.17. A flowchart showing the operation of a prediction execution unit 730e of a communication system according to a fifth embodiment of the present invention. A block diagram showing the functional configuration of a prediction unit 171f of a communication system according to a sixth embodiment of the present invention. A flowchart showing the operation of a prediction execution unit 730f of a communication system according to the sixth embodiment of the present invention. A block diagram showing the functional configuration of a prediction unit 171g of a communication system according to a seventh embodiment of the present invention. A block diagram showing the functional configurations of a switching threshold determination unit 172g and a switching determination unit 175g of a switching instruction device 17g according to the seventh embodiment of the present invention. A flowchart showing the operation of a switching flow calculation unit 177 of a switching determination unit 175g according to the seventh embodiment of the present invention. A diagram for explaining the determination of a prediction method by a prediction method determination unit in each embodiment of the present invention.

[0015] Hereinafter, a prediction device, a communication control device, and a prediction method according to embodiments of the present invention will be described in detail with reference to the drawings. Note that components having the same functions in each embodiment will be assigned the same reference numerals, and repeated explanations of those functions may be omitted.

[0016] In order to make the description of the configuration of the communication system according to the embodiment of the present invention easier to understand, an example of the configuration of a conventional communication system will be described first for comparison.

[0017] 1 is a diagram showing the configuration of a communication system 1, which is an example of a conventional communication system. The communication system 1 has a plurality of lower-level devices 3, a plurality of upper-level devices 4, and a switching instruction device 7. The lower-level devices 3 and the upper-level device 4 are connected via a network 5. The network 5 is configured to include one or more transfer devices (not shown; devices corresponding to the transfer device 15 shown in FIG. 2, which will be described later). The transfer device also functions as a switching device that can change the connection relationship between the lower-level devices 3 and the upper-level device 4.

[0018] 1, the four lower devices 3 are respectively referred to as lower devices 3-1 to 3-4, and the two upper devices 4 are respectively referred to as upper devices 4-1 and 4-2. Hereinafter, the direction from the lower device 3 to the upper device 4 will be referred to as "upstream," and the direction from the upper device 4 to the lower device 3 will be referred to as "downstream."

[0019] The lower device 3 transmits an upstream signal to the connected upper device 4. The network 5 transfers the upstream signal to the upper device 4 according to the communication path between the lower device 3 and the upper device 4. The upper device 4 also transmits a downstream signal to the connected lower device 3. The network 5 transfers the downstream signal to the lower device 3 according to the communication path between the lower device 3 and the upper device 4. The switching instruction device 7 instructs each device, such as a transfer device, to switch the communication path.

[0020] The switching instruction device 7 uses information such as traffic volume or traffic allocation volume (hereinafter collectively referred to as "traffic volume") acquired from the lower-level device 3 and the upper-level device 4 to predict in advance the traffic volume of some or all communication links (hereinafter simply referred to as "links" or "flows") in the communication path between the lower-level device 3 and the upper-level device 4. The switching instruction device 7 determines a switching threshold for each link based on the traffic volume prediction accuracy and the transmission capacity of the link.

[0021] For example, the link rate can be used as an index representing the transmission capacity of a link. The link rate is the maximum communication speed of the link. The switching threshold is a threshold used to determine whether congestion requiring a switch of the communication path (route switching) is predicted to occur. For example, even for the same link rate, the switching instruction device 7 determines a lower switching threshold as the prediction accuracy of the traffic volume becomes lower.

[0022] The switching instruction device 7 determines whether congestion will occur for each link using the predicted traffic volume and a switching threshold. If congestion is predicted to occur in any link, the switching instruction device 7 determines to perform path switching for load balancing. The switching instruction device 7 instructs switching of communication paths (path switching) so that at least a portion of the traffic transmitted through a link predicted to be congested is transmitted through a link predicted not to be congested. This enables optimal communication path switching according to prediction accuracy. Furthermore, link congestion delays are reduced, enabling higher bandwidth utilization efficiency.

[0023] As shown in FIG. 1 , the switching instruction device 7 includes a prediction unit 71, a switching threshold determination unit 72, and a switching determination unit 75. The prediction unit 71 acquires prediction information from each device to be used for predicting the traffic volume of each link. The prediction information includes, for example, the traffic volume observed by the device, the traffic allocation volume to the device, the number of terminals connected to the device, and communication quality. The prediction unit 71 predicts the future traffic volume for each link using the prediction information. Any existing technology can be used to predict the future traffic volume.

[0024] The switching threshold determination unit 72 includes a prediction accuracy calculation unit 73 and a threshold calculation unit 74. The prediction accuracy calculation unit 73 calculates prediction accuracy for each link based on past prediction results of future traffic volume and actual traffic volume. The threshold calculation unit 74 determines a switching threshold for each link such that the ratio to the link's transmission capacity (e.g., link rate) becomes lower as the prediction accuracy becomes lower.

[0025] The switching determination unit 75 determines in advance whether congestion will occur for each link using the predicted future traffic volume value and a switching threshold. If congestion is predicted, the switching determination unit 75 determines to perform route switching for load balancing. For example, the switching determination unit 75 determines to perform route switching from a link where congestion is predicted to occur to a link with a lower link utilization rate. The switching determination unit 75 instructs each device to perform processing for route switching.

[0026] An example in which the communication system 1 is applied to a mobile communication system will be described below.

[0027] 2 is a diagram showing the configuration of a mobile communication system 10, which is an example of a conventional communication system. The mobile communication system 10 is an example of the communication system 1. The mobile communication system 10 is, for example, a fifth-generation mobile communication system (5G). The mobile communication system 10 includes a plurality of terminal stations 11, a plurality of antenna stations 12, a plurality of remote stations 13, a plurality of central stations 14, a transfer device 15, a switching instruction device 17, and a resource allocation device 16.

[0028] The terminal station 11, the antenna station 12, the distributed station 13, and the central station 14 are UE (User Equipment), RU (Radio Unit), DU (Distributed unit), and CU (Central unit), respectively, of a fifth-generation mobile communication system. The distributed station 13 is an example of a lower-level device 3, and the central station 14 is an example of a higher-level device 4. However, the combination of a device corresponding to the lower-level device 3 and a device corresponding to the higher-level device 4 is not limited to the combination of the distributed station 13 and the central station 14, and is arbitrary. The transfer device 15 is an example of a transfer device that constitutes the network 5.

[0029] The transfer device 15 also functions as a switching device that can change the connection relationship between the remote stations 13 and the central station 14. The transfer function and the switching function of the transfer device 15 may be realized by separate devices. The switching instruction device 17, the transfer device 15, the resource allocation device 16, and other devices do not necessarily need to be separate devices, and may be configured as an integrated device.

[0030] The mobile communication system 10 is connected to an upper network 20. Here, the M (M is an integer equal to or greater than 1) remote stations 13 are referred to as remote stations 13-1 to 13-M, respectively. Furthermore, the Km (Km is an integer equal to or greater than 1) antenna stations 12 subordinate to the remote station 13-m (m is an integer equal to or greater than 1 and equal to or less than M) are referred to as antenna stations 12-m, respectively. Furthermore, the N (N is an integer equal to or greater than 2) central stations 14 are referred to as central stations 14-1 to 14-N, respectively. Therefore, the mobile communication system 10 shown in FIG. 2 is an example of a mobile communication system in which M=4, K1=2, K2=2, K3=2, K4=2, and N=2.

[0031] The terminal station 11 transmits and receives radio signals to and from the antenna station 12 using radio resources allocated by the remote station 13. The allocated radio resources include information indicating the start and end timings of time periods during which radio signal transmission and reception are permitted. The start and end timings are represented by, for example, slots. A slot is a unit of scheduling for data transmission and reception in a radio frame. The allocated radio resources may further include information indicating a coding rate and a modulation method.

[0032] The antenna station 12 receives uplink data from the terminal station 11 via a wireless signal. The antenna station 12-m sets the received uplink data as an uplink signal and transmits the uplink signal to the remote station 13-m via a wired interface. The antenna station 12-m also receives a downlink signal from the remote station 13-m via a wired interface. The antenna station 12 transmits the downlink data addressed to the terminal station 11, which is set in the received downlink signal, to the terminal station 11 via a wireless signal.

[0033] The remote station 13-m receives uplink signals from each of the Km antenna stations 12-m. The uplink signals received by the remote station 13-m include uplink data received by the antenna station 12-m from the terminal stations 11 under its control. The remote station 13 generates an uplink signal that aggregates the uplink data and transmits the generated uplink signal to the central station 14 to which the remote station 13 is connected. The remote station 13 also receives a downlink signal from the central station 14 to which the remote station 13 is connected, in which downlink data addressed to the terminal stations 11 under its control is set. The remote station 13-m converts the received downlink signal into a downlink signal corresponding to the radio signal to be transmitted from each antenna station 12-m. The remote station 13-m transmits the converted downlink signal to the antenna station 12-m corresponding to the downlink signal.

[0034] The central station 14 aggregates the uplink signals received from the subordinate remote stations 13 and transfers them to the upper network 20. The central station 14 also receives downlink signals from the upper network 20, in which downlink data addressed to the terminal stations 11 is set, and transfers the received downlink signals to the remote stations 13 connected to the destination terminal stations 11.

[0035] The transfer device 15 is connected to the remote stations 13, the central station 14, and the switching instruction device 17. The transfer device 15 is an example of a switching device that switches (switches) communication paths. The transfer device 15 transfers signals along the communication path between the remote stations 13 and the central station 14. That is, the transfer device 15 transfers upstream signals received from the remote stations 13 to the destination central station 14 along the communication path. The transfer device 15 also transfers downstream signals received from the central station 14 to the destination remote station 13 along the communication path. The transfer of signals along the communication path is performed under instructions from the switching instruction device 17. The resource allocation device 16 manages the resources of the central station 14.

[0036] The switching instruction device 17 is connected to the remote station 13, the central station 14, and the transfer device 15. The switching instruction device 17 may further be connected to the resource allocation device 16. The switching instruction device 17 instructs the transfer device 15 (switching device) to switch the communication path between the remote station 13 and the central station 14. In the mobile communication system 10, the communication path between the remote station 13 and the central station 14 is determined based on the connection relationship between the remote station 13 and the central station 14. Therefore, in the mobile communication system 10, the path between the remote station 13 and the central station 14 is switched by controlling the transfer device 15 (switching device) to change the central station 14 to which the remote station 13 is connected.

[0037] Also, as an example, it is assumed here that the switching instruction device 17 of the mobile communication system 10 determines, based on the upstream traffic, whether or not congestion has occurred on the central station 14 side, which corresponds to the higher-level device 4. Therefore, unless otherwise specified, traffic in the following description refers to the traffic of upstream signals.

[0038] 3 is a block diagram showing the configuration of a switching instruction device 17 of a mobile communication system 10, which is an example of a conventional communication system. The switching instruction device 17 includes a prediction unit 171, a switching threshold value determination unit 172, and a switching determination unit 175.

[0039] The prediction unit 171 receives radio information from the remote stations 13 as prediction information. The radio information includes information indicating radio resources to be allocated to terminal stations 11 under the control of the remote stations 13. The radio information is, for example, link control information (DCI: Downlink Control Information). The prediction unit 171 uses the radio information from the remote stations 13 to calculate the future traffic volume of each remote station 13.

[0040] Furthermore, the prediction unit 171 calculates the future traffic volume of each upper link based on the connection relationship between the remote stations 13 and the aggregation station 14 and the future traffic volume of each remote station 13. The upper link is the link between the aggregation station 14 and the transfer device 15. The upstream traffic volume of the upper link between the aggregation station 14-n (n is an integer between 1 and N) and the transfer device 15 corresponds to the total value of the traffic volume of the upstream signals that the aggregation station 14-n receives from each of the remote stations 13 under its control. The prediction unit 171 outputs the future traffic volume of the upper link to the switching threshold determination unit 172 and the switching determination unit 175, respectively.

[0041] The switching threshold determination unit 172 includes a prediction accuracy calculation unit 173 and a threshold calculation unit 174. The prediction accuracy calculation unit 173 receives information indicating actual traffic volume from the remote station 13. The prediction accuracy calculation unit 173 calculates the actual traffic volume flowing through each upper link based on the actual traffic volume and the connection relationship between the remote station 13 and the central station 14. The prediction accuracy calculation unit 173 calculates prediction accuracy for each upper link based on the future traffic volume predicted in the past by the prediction unit 171 and the actual traffic volume. The threshold calculation unit 174 calculates a switching threshold for each upper link based on the link rate and prediction accuracy. The threshold calculation unit 174 outputs the calculated switching threshold for each upper link to the switching determination unit 175.

[0042] The switching determination unit 175 calculates a predicted link utilization rate for each upper link based on the future traffic volume value acquired from the prediction unit 171 and the switching threshold acquired from the threshold calculation unit 174. If there is an upper link for which it is determined that congestion will occur due to a high predicted link utilization rate, the switching determination unit 175 determines to perform path switching by changing the aggregation station 14 to which the remote station 13 is connected. If the switching determination unit 175 determines to perform path switching, it instructs each device (the remote station 13, the aggregation station 14, and the transfer device 15) to perform path switching.

[0043] Specifically, the switching determination unit 175 instructs the central station 14 that is the connection destination before the path switching (hereinafter referred to as the "switching source") to release the connection with the remote station 13, and instructs the central station 14 that is the connection destination after the path switching (hereinafter referred to as the "switching destination") to connect with the remote station 13. Furthermore, the switching instruction device 17 instructs the transfer device 15 (switching device) to switch the path so that signals are transferred via the communication path after the path switching.

[0044] 4 is a diagram showing a processing flow in the switching instruction device 17 of the mobile communication system 10, which is an example of a conventional communication system. The prediction unit 171 predicts future traffic volume of an upper link, for example, for a period D(a) (step S1). The prediction accuracy calculation unit 173 obtains information indicating the future traffic volume predicted in step S1 from the prediction unit 171 and stores the information until it obtains information indicating the actual traffic volume for the period D(a) (step S2).

[0045] The prediction accuracy calculation unit 173 receives information indicating the actual traffic volume of the upper link for period D(a) (step S3). The prediction accuracy calculation unit 173 calculates the prediction accuracy based on the stored future traffic volume for period D(a) and the received actual traffic volume for period D(a). The threshold calculation unit 174 determines a switching threshold based on the link rate and prediction accuracy of the upper link (step S4). The threshold calculation unit 174 outputs the determined switching threshold to the switching determination unit 175.

[0046] Meanwhile, the prediction unit 171 calculates the future traffic volume of the upper link for period D(b), which is a period after period D(a) (step S5). The switching determination unit 175 acquires the switching threshold output from the threshold calculation unit 174 in step S4 (step S6). Furthermore, the switching determination unit 175 acquires information indicating the future traffic volume for period D(b), calculated in step S5, from the prediction unit 171 (step S7). The switching determination unit 175 determines whether or not route switching is necessary based on the acquired switching threshold and the value of the future traffic volume for period D(b) (step S8).

[0047] As described above, the future traffic volume acquired by the switching threshold value determination unit 172 is the traffic volume at a time earlier than the future traffic volume acquired by the switching determination unit 175 .

[0048] Next, a detailed description will be given of the processing of the switching instruction device 17. First, a description will be given of the processing of predicting future traffic volume of each upper link in the prediction unit 171. The prediction unit 171 performs the processing of steps S1 and S5 in Fig. 4 by the processing of Figs. 5 to 8 below.

[0049] 5 is a flowchart showing the process of predicting the traffic volume of a remote station in the prediction unit 171. At time t(i-1), the remote station 13 notifies the terminal station 11 of DCI(i) for the next transmission period, and further notifies the switching instruction device 17 of DCI information in which DCI(i) is set (i is an integer). DCI(i) is DCI that indicates the radio resources to be allocated to the terminal station 11 in slot SL(i) from time t(i) to time t(i+1), and the coding rate and modulation scheme to be used by the terminal station 11.

[0050] Specifically, the DCI includes scheduling information, an MCS (Modulation and Coding Scheme), and the like, which are necessary for the terminal station 11 to transmit uplink data. The scheduling information is represented by a resource block. The resource block is represented by a channel and the transmission start timing and transmission end timing using that channel. The MCS represents the data modulation and channel coding rate. The TBS (Transport Block Size) is determined based on the resource block size and the MCS. The TBS makes it possible to predict the traffic volume.

[0051] At time t(i-1), the switching instruction device 17 receives DCI information in which DCI(i) is set from each remote station 13. Based on the information included in DCI(i), the prediction unit 171 calculates the predicted traffic volume DU(i) of the remote station 13 for the period D(i) corresponding to the slot SL(i) (step S111). Furthermore, the prediction unit 171 calculates the allocated time interval T(i) from the difference between time t(i) and time t(i+1) (step S112). The allocated time interval T(i) corresponds to the duration of the slot SL(i), i.e., the duration of the period D(i) from time t(i) to time t(i+1). The prediction unit 171 may receive the allocated time interval T(i) from the remote station 13. The prediction unit 171 stores the predicted traffic volume DU(i) of each remote station 13, the allocated time interval T(i), and the period D(i) in association with each other.

[0052] The prediction unit 171 creates a traffic volume prediction model (step S113). The traffic volume prediction model is a model that uses traffic volumes of q (e.g., q=100) slots in a time series as input data and predicts the traffic volume of the slot next to the input data. Details of creating the traffic volume prediction model will be described later with reference to FIG. 6.

[0053] The prediction unit 171 predicts the future traffic volume DU(i+1) of each remote station 13 during the period D(i+1) from time t(i+1) to time t(i+2) using the traffic volume prediction model created in step S113 (step S114). Details of the prediction process will be described later with reference to Fig. 7. After the process of step S114, the prediction unit 171 performs the process of Fig. 8, which will be described later, to calculate the traffic volume for each upper link.

[0054] 6 is a flowchart showing the process of creating a traffic volume prediction model by the prediction unit 171. FIG. 6 shows the detailed process of step S113 in FIG.

[0055] The prediction unit 171 stores in advance K time intervals to be processed: T_1, T_2, ..., T_k-1, and T_req. The time intervals T_1 to T_k-1 are the time lengths of slots available for allocation of resource blocks to the terminal station 11. For example, T_1 to T_k-1 are 125 [us], 250 [us], 500 [us], and 1 [ms]. The allocation time interval T(i) is the same value as one of the time intervals T_1 to T_k-1. The time interval T_req is a value required by an application or the like. The time interval T_req is a value greater than T_1 to T_k-1. The prediction unit 171 selects one of the time intervals T_1 to T_req that has not yet been selected, and sets it as T_x (step S121).

[0056] The prediction unit 171 determines whether the allocated time interval T(i) is equal to or greater than the time interval T_x (step S122). If the prediction unit 171 determines that the allocated time interval T(i) is equal to or greater than the time interval T_x (step S122: YES), the prediction unit 171 converts the predicted traffic volume DU(i) of the remote station 13 into a predicted traffic volume for each time interval T_x (step S123).

[0057] That is, the prediction unit 171 calculates that a predicted traffic volume DU_T_x of predicted traffic volume DU(i) × (time interval T_x / allocated time interval T(i)) will occur for each time interval T_x between time t(i) and time t(i+1). For example, assume that the allocated time interval T(i) is 250 us and the time interval T_x is 125 us. The prediction unit 171 predicts that a traffic volume of DU_125 us = DU(i) / 2 will occur between time t(i) and time t(i)+125 us and between time t(i)+125 us and time t(i+1), respectively.

[0058] The prediction unit 171 adds the predicted traffic volume calculated in step S123 to the time-series predicted traffic volume previously calculated for the time interval T_x for each remote station 13 and stores the result (step S124). The prediction unit 171 learns a traffic volume prediction model for the time interval T_x using the stored time-series predicted traffic volumes of the remote stations 13 for the time interval T_x (step S125).

[0059] For example, the time-series predicted traffic volumes of the remote station 13 for the time interval T_x are assumed to be DU_T_x(P), DU_T_x(P-1), DU_T_x(P-2), ..., DU_T_x(1) in order of most recent time (P is an integer equal to or greater than 2). The prediction unit 171 generates learning data for each remote station 13, using DU_T_x(p) as the correct output data and subtracting 1 from the value of p in order from P, with DU_T_x(p-1) to DU_T_x(p-q) as input data (q is an integer equal to or greater than 1, for example, q=100). The prediction unit 171 uses this generated learning data to train a traffic volume prediction model that represents the correspondence between input data and output data.

[0060] On the other hand, if the prediction unit 171 determines in step S122 that the allocated time interval T(i) is smaller than the time interval T_x (step S122: NO), it performs the process of step S126. The prediction unit 171 records the predicted traffic volume DU(i) of the remote station 13-m in a buffer within the prediction unit 171 that corresponds to the remote station 13-m and the time interval T_x (step S126). The prediction unit 171 determines whether the predicted traffic volume for the time interval T_x has been recorded in the buffer that corresponds to each remote station 13 and the time interval T_x (step S127).

[0061] For example, assume that the allocated time interval T(i) is 250 [us] and the time interval T_x is 500 [us]. The prediction unit 171 determines whether the predicted traffic volume for T_x / T(i) = 2 has been recorded in the buffer corresponding to 500 [us]. If the prediction unit 171 determines that the predicted traffic volume for the time interval T_x has not been recorded in the buffer corresponding to the time interval T_x (step S127: NO), the prediction unit 171 performs the processing from step S128.

[0062] The prediction unit 171 performs the process of step S124 when it determines that the predicted traffic volume for the time interval T_x has been recorded in the buffer corresponding to the time interval T_x for each remote station 13 (step S127: YES). That is, the prediction unit 171 reads out the predicted traffic volume DU(i-T_x / T(i)+1) to the predicted traffic volume DU(i) for each remote station 13 from the buffer corresponding to the time interval T_x, and sums them up to obtain DU_T_x(P).

[0063] For example, suppose the allocated time interval T(i) is 250 [us] and the time section T_x is 500 [us]. The prediction unit 171 sums the predicted traffic volume DU(i-1) and the predicted traffic volume DU(i) to calculate DU_500 [us] (P).

[0064] The prediction unit 171 stores the time-series predicted traffic volumes DU_T_x(1) to DU_T_x(P-1) of each remote station 13 that were previously calculated for the time interval T_x. The prediction unit 171 adds the newly calculated predicted traffic volume DU_T_x(P) to these predicted traffic volumes DU_T_x(1) to DU_T_x(P-1) for each remote station 13 and stores them (step S124). The prediction unit 171 performs the process of step S125 and learns a traffic volume prediction model using the time-series predicted traffic volume for the time interval T_x.

[0065] After the process of step S125, or if the determination in step S127 is NO, the prediction unit 171 determines whether all of the K time intervals to be processed, T_1 to T_req, have been selected (step S128). If there is an unselected time interval (step S128: NO), the prediction unit 171 returns to step S121 and sets the newly selected time interval to T_x.

[0066] Then, if the prediction unit 171 determines that all K time intervals to be processed have been selected (step S128: YES), it stores a traffic volume prediction model corresponding to each of the K time intervals T_1 to T_req and a predicted time, which is the time from when input data is input to the traffic volume prediction model to when output data is output (step S129).

[0067] Fig. 7 is a flowchart showing the process of predicting future traffic volume by the prediction unit 171. Fig. 7 shows detailed processing in step S114 of Fig. 5. The prediction unit 171 determines whether it is possible to predict future traffic volume by time t(i+1), when the next period D(i) ends (step S131). Specifically, the prediction unit 171 determines whether the sum of the allocated time interval T(i) and the switching time Tc is longer than the prediction time of the prediction model for the traffic volume of the allocated time interval T(i).

[0068] If the allocation time interval T(i) + switching time Tc is longer than the predicted time (T(i) + Tc > predicted time), the prediction unit 171 determines that it is possible to predict future traffic volume by the end of the next period D(i) (step S131: YES). The prediction unit 171 inputs the p most recent time-series predicted traffic volumes DU(P) to DU(P-p+1) for the allocation time interval T(i) into the traffic volume prediction model created for each remote station 13. The prediction unit 171 calculates the future traffic volume DU(i+1) for each remote station 13 during the period D(i+1) from time t(i+1) to time t(i+2) (step S132). Note that t(i+2) = t(i+1) + T(i). The prediction unit 171 stores the calculated future traffic volume DU(i+1) for each remote station 13.

[0069] If the predicted time is equal to or greater than the allocation time T(i) + switching time Tc (T(i) + Tc ≦ predicted time), the prediction unit 171 determines that it is impossible to predict future traffic volume by the end of the next period D(i) (step S131: NO). The prediction unit 171 performs prediction from time t(i+1) to the time interval T_req (e.g., 5 ms) ahead, which satisfies the service request delay. For each remote station 13, the prediction unit 171 adds up the predicted traffic volumes of the time series for the allocation time interval T(i) in units of T_req / T(i) in descending order of time to calculate the predicted traffic volume of the time series for the time interval T_req. For each remote station 13, the prediction unit 171 inputs the predicted traffic volumes of the most recent q time series for the time interval T_req into the traffic volume prediction model created for the time interval T_req.

[0070] As a result, the prediction unit 171 calculates the future traffic volume DUreq(i+1) of each remote station 13 during the period D(i+1) from time t(i+1) to time t(i+2) (step S133). Note that t(i+2) = t(i+1) + T_req. The prediction unit 171 sets the future traffic volume DUreq(i+1) to the future traffic volume DU(i+1) calculated in step S114 of FIG. 5 .

[0071] FIG. 8 is a flowchart showing the process of predicting the traffic volume of the upper link in the prediction unit 171. After the process shown in the flowchart of FIG. 5, the prediction unit 171 executes the process of predicting the traffic volume of the upper link shown in the flowchart of FIG. 8. The traffic volume of the upper link between the central station 14-n and the transfer device 15 is the same as the traffic volume of the central station 14-n. Therefore, the prediction unit 171 first acquires bearer information that the switching instruction device 17 has received in advance from the central station 14 (step S141). The bearer information is used as connection information that indicates the connection between the upper device and the lower device. By acquiring the bearer information, the prediction unit 171 recognizes the connection relationship between the remote station 13 and the central station 14.

[0072] The prediction unit 171 sums up, for each central station 14, the future traffic volume DU(i+1) of the subordinate remote stations 13 calculated in step S114 of Fig. 5. As a result, the prediction unit 171 calculates the future traffic volume L(i+1), which is the traffic volume of the upper link connected to the central station 14 during the period D(i+1) (step S142). Specifically, the prediction unit 171 calculates the future traffic volume L(i+1) as follows:

[0073] Future traffic volume L(i+1)=Σfuture traffic volume DU(i+1)×if(aggregate station-distributed station connection)

[0074] The above if (aggregate station-remote station connection) is a flag that takes the value of 1 if the aggregate station 14 is designated as the connection destination of the remote station 13, and takes the value of 0 if not. The prediction unit 171 outputs the period D(i+1) and the future traffic volume L(i+1) of each upper link to the switching threshold determination unit 172 and the switching determination unit 175 (step S143).

[0075] 9 is a flowchart showing the calculation process of the switching threshold value in the switching threshold value determination unit 172. The process shown in the flowchart in FIG. 9 corresponds to the process in step S4 in FIG. 4. The prediction accuracy calculation unit 173 receives information on the period D(a) and the future traffic volume L(a) of each upper link from the prediction unit 171 (step S151). This corresponds to the period D(i+1) and the future traffic volume L(i+1) of each upper link output by the prediction unit 171 in step S143 in FIG. 8.

[0076] Next, the prediction accuracy calculation unit 173 acquires the actual traffic volume of each remote station 13 during the period D(a) (step S152). For each central station 14, the prediction accuracy calculation unit 173 sums up the actual traffic volumes for the period D(a) acquired in step S152 for the remote stations 13 under its control, to calculate the actual traffic volume L'(a), which is the actual traffic volume of the upper link connected to the central station 14. Note that the prediction accuracy calculation unit 173 may receive the actual traffic volume L'(a) during the period D(a) from the central station 14.

[0077] The prediction accuracy calculation unit 173 acquires the link rate of each upper link. The prediction accuracy calculation unit 173 calculates the prediction accuracy for each upper link in the period D(a) (step S153). For example, the prediction accuracy calculation unit 173 calculates the following error rate as the prediction accuracy:

[0078] Error rate = | future traffic volume L(a) - actual traffic volume L'(a) | ÷ (actual traffic volume L'(a)) × 100

[0079] Alternatively, the prediction accuracy calculation unit 173 may calculate the following RMSE (Root Mean Squared Error) or MAE (Mean Absolute Error) used to evaluate prediction accuracy. Note that in the following equations (1) and (2), fi represents the future traffic volume L(a), and yi represents the actual traffic volume L'(a).

[0080]

[0081]

[0082] Next, the threshold calculation unit 174 calculates a switching threshold to be used after the period D(a) using the prediction accuracy calculated in step S153 (step S154). For example, when the threshold calculation unit 174 calculates the error rate in step S153, it calculates the switching threshold TH of the upper link by the following formula (3).

[0083] Switching threshold TH = link rate × (100 - error rate) ÷ 100 (3)

[0084] Alternatively, when the threshold calculation unit 174 calculates the RMSE or MAE by the above formula (1) or formula (2) in step S153, it converts the pre-stored lookup table into an error rate.

[0085] 10 is a diagram showing an example of a lookup table. The lookup table is data that associates prediction accuracy with an error rate and a calculation formula for a switching threshold. The threshold calculation unit 174 reads out an error rate corresponding to the prediction accuracy from the lookup table. The threshold calculation unit 174 uses the read error rate to calculate the switching threshold TH for each upper link according to the above formula (3).

[0086] In FIG. 9, the threshold calculation unit 174 outputs the switching threshold TH of each upper link calculated in step S154 to the switching determination unit 175 (step S155).

[0087] For example, accuracy is used to calculate the prediction accuracy. However, it is uncertain whether the prediction accuracy can actually be guaranteed. In this embodiment, the traffic volume is predicted in advance, and then the actual traffic flows. Therefore, as described above, the prediction accuracy is calculated using the difference between the actual traffic volume and the predicted traffic volume.

[0088] Fig. 11 is a flowchart showing the switching determination process in the switching determination unit 175. The process in Fig. 11 corresponds to the process in step S8 in Fig. 4. The switching determination unit 175 performs the process shown in Fig. 11 for each upper link. The switching determination unit 175 receives information on the period D(b) and the future traffic volume L(b) of each aggregation station 14 from the prediction unit 171 (step S161). These correspond to the period D(i+1) and the future traffic volume L(i+1) of each aggregation station 14 output by the prediction unit 171 in step S143 in Fig. 8.

[0089] The switching determination unit 175 receives the switching threshold TH for each upper link output by the threshold calculation unit 174 in step S155 of Fig. 9 (step S162). Using the switching threshold TH, the switching determination unit 175 calculates a predicted link usage rate for each upper link by the following equation (4) (step S163).

[0090] Predicted link utilization rate=future traffic volume L(b) / switching threshold TH (4)

[0091] Before substituting the future traffic volume L(b) into equation (4), the switching determination unit 175 converts it into the same traffic volume per unit time as the link rate by, for example, dividing it by the length of the period D(b). The switching determination unit 175 determines whether the calculated predicted link utilization rate exceeds 1 (step S164). If the switching determination unit 175 determines that the link utilization rate is 1 or less (i.e., if the switching determination unit 175 determines that the future traffic volume is equal to or less than the switching threshold), congestion is predicted not to occur, and therefore the processing shown in the flowchart of FIG. 11 is terminated. On the other hand, if the switching determination unit 175 determines that the predicted link utilization rate exceeds 1 (i.e., if the switching determination unit 175 determines that the future traffic volume will exceed the switching threshold), congestion is predicted to occur, and therefore route switching is to be performed (step S165).

[0092] Fig. 12 is a flowchart showing the switching determination process in the switching determination unit 175. When it is determined in step S165 of Fig. 11 that route switching is to be performed, the switching determination unit 175 starts executing the process shown in the flowchart of Fig. 12.

[0093] The switching determination unit 175 acquires from the prediction unit 171 the future traffic volume for each upper link for period D(b), information on the distributed stations 13 subordinate to the aggregate station 14 connected to that upper link, and the future traffic volume DU(b) for that distributed station 13 for period D(b). The switching determination unit 175 determines to switch the connection destination of one or more distributed stations 13 subordinate to the aggregate station 14 using an upper link whose predicted link utilization rate exceeds 1 to another aggregate station 14. At this time, the switching determination unit 175 uses the future traffic volume DU(b) of each distributed station 13 to determine the switching destination aggregate station 14 for the distributed station 13 to be switched to, so that the future traffic volume for each upper link is equal to or less than the switching threshold TH for that upper link (step S171).

[0094] The switching determination unit 175 transmits a path addition instruction to the source central station 14 to add traffic from the remote station 13 to be switched (step S172). Furthermore, the switching determination unit 175 transmits a path deletion instruction to the source central station 14 to delete traffic from the remote station 13 to be switched (step S173). Furthermore, the switching determination unit 175 transmits a path switching instruction to the transfer device 15 to instruct it to switch the communication path between the remote station 13 to be switched and the source central station 14 to the communication path between the remote station 13 to be switched and the destination central station 14 (step S174).

[0095] As described above, the communication system 1 and the mobile communication system 10, which are examples of conventional communication systems, predict future traffic volume of the upper link during period D(a) and calculate the prediction accuracy based on the future traffic volume and the actual traffic volume. The communication system 1 and the mobile communication system 10 determine the switching threshold based on the link rate and prediction accuracy of the upper link. Then, the communication system 1 and the mobile communication system 10 predict future traffic volume of the upper link during period D(b), which is a period after period D(a), and determine whether or not route switching is necessary based on the switching threshold and the future traffic volume.

[0096] However, in an actual communication environment where multiple flows each having various traffic characteristics coexist, predicting the future traffic volume of all flows collectively results in low prediction accuracy. On the other hand, predicting the future traffic volume for each flow individually increases the computational load involved in the prediction, and it may become impossible to complete route switching before congestion actually occurs. Thus, in conventional communication systems, even if a decision to perform route switching is made based on predicted traffic volumes, congestion may not be prevented, resulting in communication delays.

[0097] In contrast to this, the communication system according to each embodiment of the present invention described below can reduce the amount of calculation required for predicting future traffic volume so that route switching can be completed before congestion occurs.

[0098] Specifically, the communication system in each of the following embodiments is configured to group (classify) multiple flows to be predicted and determine a method for predicting future traffic volume for each group. Furthermore, the communication system in each of the following embodiments is configured to predict future traffic volume using linear prediction, a prediction method with a relatively low computational load, for flows that are difficult to predict using machine learning. Furthermore, the communication system in each of the following embodiments is configured to not predict future traffic volume for flows that can be predicted using machine learning but have no change in the number of terminals, for example, and to use the most recent traffic volume value as the estimation result.

[0099] By adopting such a configuration, the communication system according to each embodiment of the present invention described below can reduce the amount of calculation required for predicting future traffic volume. As a result, the communication system according to each embodiment can suppress congestion and reduce communication delays while satisfying communication quality, even in a general communication environment where multiple flows coexist.

[0100] <First Embodiment> A communication system according to a first embodiment of the present invention will now be described. Note that the overall configuration of the communication system according to this embodiment is similar to the configuration of the conventional communication system 1 shown in the overall configuration diagram of Figure 1 described above, and therefore a description thereof will be omitted. Furthermore, the overall configuration of a mobile communication system in which the communication system according to this embodiment is applied to mobile communication is also similar to the configuration of the conventional mobile communication system 10 shown in the overall configuration diagram of Figure 2 described above, and therefore a description thereof will be omitted.

[0101] Furthermore, the block diagram showing the functional configuration of the switching instruction device of the communication system in the first embodiment is the same as the block diagram showing the functional configuration of the switching instruction device 17 of the conventional communication system 1 shown in FIG. 3 described above. Hereinafter, the switching instruction device in the first embodiment will be referred to as switching instruction device 17a. The switching instruction device 17a is, for example, an information processing device such as a general-purpose computer. The switching instruction device 17a in this embodiment and the conventional switching instruction device 17 described above differ in the configuration of the prediction unit. The configuration of the prediction unit 171a of the communication system in this embodiment will be described below.

[0102] 13 is a block diagram showing the functional configuration of the prediction unit 171a of the communication system according to the first embodiment of the present invention. As shown in Fig. 13, the prediction unit 171a includes a flow classification unit 710, a flow information storage unit 720, a prediction execution unit 730, and a total traffic calculation unit 740.

[0103] The flow classification unit 710 classifies flows input to the network 5 within a predetermined period (e.g., 1 millisecond or 1 second) based on traffic characteristics. Here, a flow is defined as a flow transmitted from a lower-level device 3 to a higher-level device 4 via the network 5.

[0104] Here, flows with the same traffic pattern are defined as one flow. Flows with the same traffic pattern are, for example, flows with the same QoS. Furthermore, even if flows are sent from the same lower-level device 3 and upper-level device 4, if they have different traffic patterns (for example, different QoS), they are defined as different flows.

[0105] Fig. 14 is a schematic diagram for explaining the definition of a flow, which is a prerequisite for the first embodiment of the present invention. Fig. 14 illustrates an example in which there are three terminals (UE) that are lower-level devices 3, and the thick circles represent the flows from each terminal. The flows from each terminal may include multiple flows with different traffic patterns (different QoS).

[0106] The thick cylinder in the top row shows two traffic pattern flows represented by thin cylinders with light hatching and thin cylinders with medium hatching, while the thick cylinder in the middle row shows three traffic pattern flows represented by thin cylinders with light hatching, thin cylinders with medium hatching, and thin cylinders with dark hatching, respectively. The thick cylinder in the bottom row shows only one traffic pattern flow represented by a thin cylinder with light hatching.

[0107] For example, a thin cylinder with a dark hatch represents video traffic with a frame rate of 60 fps (frames per second), a thin cylinder with a medium hatch represents the flow of video traffic with a frame rate of 30 fps, and a thin cylinder with a light hatch represents the flow of file transfer traffic via FTP (File Transfer Protocol), etc.

[0108] Here, it is assumed that flows with the same traffic pattern (with the same QoS) have the same traffic characteristics (for example, traffic periodicity, etc.) The prediction unit 171a classifies flows by traffic pattern and predicts future traffic volume.

[0109] The flow classification unit 710 classifies flows by flow type based on information defined in, for example, 5QI (5G QoS Identifier) ​​or S-NSSAI (Sub Network Slice Selection Assist Information), etc. The flow classification unit 710 calculates the total number of terminals and traffic volume for each classified flow and stores them in a flow information table stored in the flow information storage unit 720.

[0110] The flow information storage unit 720 is configured to include a storage medium, for example, a semiconductor memory such as RAM (Random Access Memory), EEPROM (Electrically Erasable Programmable Read Only Memory) and ROM (Read Only Memory), a flash memory such as SSD (Solid State Drive), a magnetic disk such as HDD (Hard Disk Drive), an optical disk, or any combination of these storage media (not shown).

[0111] 15 is a diagram showing an example of the configuration of a flow information table in the first embodiment of the present invention. As shown in Fig. 15, for example, the flow information table is tabular data in which a flow ID (Identifier), a flow type, a traffic volume, the number of terminals, and the number of flows are associated with each other.

[0112] The flow ID is an identifier that identifies a classified flow. The flow type is a value that represents the flow characteristics. In the flow information table illustrated in FIG. 15, the value of the flow type is a 5QI value. The traffic volume is the total value of the traffic volume for each classified flow. In the flow information table illustrated in FIG. 15, the unit of traffic volume is kilobytes. The number of terminals is the total value of the number of terminals for each classified flow. In the flow information table illustrated in FIG. 15, the unit of traffic volume is units.

[0113] The prediction execution unit 730 determines whether future traffic volume can be predicted using a prediction method with a relatively low computational complexity (hereinafter referred to as "low computational complexity prediction") for each flow grouped (classified) by the flow classification unit 710. For flows determined to be predictable using low computational complexity prediction, the prediction execution unit 730 predicts future traffic volume using the low computational complexity prediction. For flows determined to be unpredictable using low computational complexity prediction, the prediction execution unit 730 predicts future traffic volume using a prediction method with a relatively high computational complexity (hereinafter referred to as "high computational complexity prediction"). The prediction execution unit 730 outputs the value of the future traffic volume predicted for each flow to the total traffic calculation unit 740.

[0114] Low-computational-amount prediction is a simpler prediction method that can predict future traffic volume with a relatively smaller amount of calculation compared to high-computational-amount prediction. The low-computational-amount prediction here refers to a prediction method using linear prediction, for example. The high-computational-amount prediction here refers to a prediction method using machine learning, for example.

[0115] In contrast to the conventional switching instruction device 17 that performs high-computational-amount prediction for all flows to be predicted, the switching instruction device 17a in this embodiment is configured to use low-computational-amount prediction for flows that can use low-computational-amount prediction, thereby enabling the switching instruction device 17a in this embodiment to reduce the amount of calculation required to predict future traffic volume.

[0116] As shown in FIG. 13, the prediction execution unit 730 includes a prediction method determination unit 731 , a large calculation amount prediction unit 732 , and a small calculation amount prediction unit 733 .

[0117] The prediction method determination unit 731 determines whether or not it is possible to predict future traffic volume using low-computational-amount prediction, based on the traffic volume and traffic pattern, for each flow grouped by the flow classification unit 710. More specifically, the prediction method determination unit 731 calculates the time-series fluctuation of traffic volume (traffic volume spread (burstiness)) for each flow grouped by the flow classification unit 710, and determines whether or not it is possible to predict future traffic volume using low-computational-amount prediction, based on the calculation result.

[0118] The prediction method determination unit 731 may refer to the information stored in the flow information table shown in FIG. 15, identify the traffic volume and traffic pattern based on the flow ID of each flow, and determine whether prediction using low-computational-amount prediction is possible. In this case, the prediction method determination unit 731 identifies the flow ID based on, for example, information related to 5QI or S-NSSAI stored in a storage medium (not shown) included in the prediction unit 171a. The information related to 5QI or S-NSSAI may be stored in the flow information storage unit 720. The storage medium (not shown) may include, for example, a storage medium such as a RAM, a semiconductor memory such as an EEPROM or a ROM, a flash memory such as an SSD, a magnetic disk such as an HDD, an optical disk, or any combination of these storage media.

[0119] If the prediction method determination unit 731 determines that future traffic volume can be predicted using low-computational-amount prediction, it outputs information indicating the flow ID of the determined flow and the current traffic volume to the low-computational-amount prediction unit 733. On the other hand, if the prediction method determination unit 731 determines that future traffic volume cannot be predicted using low-computational-amount prediction, it outputs information indicating the flow ID of the determined flow and the current traffic volume to the high-computational-amount prediction unit 732.

[0120] The computationally intensive prediction unit 732 predicts future traffic volume using the computationally intensive prediction and outputs the value of the future traffic volume predicted for each flow to the total traffic calculation unit 740.

[0121] The small calculation amount prediction unit 733 predicts future traffic volume using the small calculation amount prediction, and outputs the value of the future traffic volume predicted for each flow to the total traffic calculation unit 740.

[0122] The following will explain in more detail, with examples, the processing performed by the prediction execution unit 730. Fig. 16 is a flowchart showing the operation of the prediction execution unit 730 in the communication system according to the first embodiment of the present invention.

[0123] The prediction method determination unit 731 of the prediction execution unit 730 calculates the maximum value (Traffic_burst) of the difference between the immediately preceding traffic volume and the current traffic volume for each flow classified by the flow classification unit 710. The prediction method determination unit 731 calculates the difference in traffic volume (burstiness) based on the calculated maximum value (Traffic_burst) (step S181).

[0124] The prediction method determination unit 731 may calculate the difference in traffic volume (burstiness) based on the traffic volume at the time of traffic transmission and the 5QI (or S-NSSAI) that can be obtained from the lower-level device 3. Alternatively, the prediction method determination unit 731 may obtain information indicating the difference in traffic volume (burstiness) from a controller (not shown) of the lower-level device 3.

[0125] The prediction method determination unit 731 may calculate the difference in traffic volume (burstiness) based on a traffic allocation volume (TBS) that is a scheduling result for a wireless section instead of the traffic volume. Alternatively, the prediction method determination unit 731 may calculate the difference in traffic volume (burstiness) based on the traffic volume (or traffic allocation volume) at the time of traffic transmission that can be obtained from a SW (Switch) or RT (Remote Terminal) installed between devices (for example, between the lower device 3 and the upper device 4) and the DSCP.

[0126] The configuration of the communication system in this embodiment is not limited to a mobile network, but can also be applied to a wireless network, etc. In this case, the prediction method determination unit 731 may acquire information indicating the traffic volume from a Wi-Fi (registered trademark) access point, etc.

[0127] Next, the prediction method determination unit 731 determines whether the difference between the calculated traffic volumes is small. Specifically, the prediction method determination unit 731 determines whether the maximum value (Traffic_burst) of the difference between the previous traffic volume and the current traffic volume calculated in step S181 is smaller than a predetermined threshold (T_th) (step S182).

[0128] If it is determined that the difference in traffic volume is small (step S182, Yes), the prediction method determination unit 731 determines that the future traffic volume can be predicted using the low-computational-amount prediction. On the other hand, if it is determined that the difference in traffic volume is not small (step S182, No), the prediction method determination unit 731 determines that the future traffic volume cannot be predicted using the low-computational-amount prediction.

[0129] A specific example will be described below. For example, it is assumed that the time-series fluctuation (traffic1_1, traffic1_2, . . . ) of the traffic volume (Traffic1) of flow 1 is as shown in the following equation (5).

[0130] Traffic1=[traffic1_1, traffic1_2,...]=[200,200,...]...(5)

[0131] At this time, traffic_burst_1_2, which is the difference between traffic1_1 and traffic1_2, is calculated as in the following equation (6).

[0132] traffic_burst_1_2=abs(traffic1_2-traffic1_1)=200-200=0...(6)

[0133] The sum of the difference values ​​between traffic1_1 and each of the subsequent traffic volumes (traffic1_2+traffic1_3+ . . . +traffic1_N) is expressed by the following equation (7).

[0134] Traffic_bust_1=sum(traffic_bust1_2+traffic_bust1_3+...+traffic_bust1_N)...(7)

[0135] Here, the value of N may be, for example, the maximum number of time-series data of traffic volume stored in the flow information storage unit 729, or may be any number (for example, N = 100). For example, when N = 10, the sum of the difference values ​​between traffic1_1 and each of the subsequent traffic volumes is calculated as shown in the following formula (8) if the traffic volume value is the value of flow 1 recorded in the flow information table illustrated in FIG.

[0136] Traffic_bust_1=sum(traffic_bust1_2+traffic_bust1_3+...+traffic_bust1_10)=100...(8)

[0137] The prediction method determination unit 731, for example, compares the calculated value of Traffic_bust_1 with a predetermined threshold (T_th) to determine the magnitude of the difference (burstiness) in the traffic volume of flow 1. The threshold T_th may be set to a predetermined fixed value, or may be set to a variable value such as a predetermined percentage of the traffic volume (e.g., 80% of the maximum traffic volume of each flow).

[0138] When the difference in traffic volume is small, it is possible to predict future traffic by low-computation-amount prediction using, for example, linear prediction, etc. Therefore, when the difference in traffic volume is small, the prediction method determination unit 731 causes the low-computation-amount prediction unit 733 to execute low-computation-amount prediction using, for example, linear prediction, etc. (step S183).

[0139] On the other hand, if the difference in traffic volume is not small, it is difficult to predict future traffic using low-computational-volume prediction, for example, linear prediction, etc. Therefore, if the difference in traffic volume is not small, the prediction method determination unit 731 causes the high-computational-volume prediction unit 732 to perform high-computational-volume prediction using, for example, machine learning (step S184).

[0140] When the value of Traffic_burst exceeds the threshold T_th as described above, the large-computation-amount prediction unit 732 performs large-computation-amount prediction (time-series prediction). Therefore, in this case, the large-computation-amount prediction unit 732 first calculates the prediction period. Here, the value of the threshold T_th is set to, for example, 80% of the maximum traffic volume of each flow.

[0141] For example, in the case of flow 1 listed in the flow information table shown in Fig. 15, the maximum traffic volume is 400 kilobytes. Therefore, the threshold value (T_th_1) for flow 1 is set as shown in the following equation (9).

[0142] T_th_1=400×0.8=320…(9)

[0143] For example, as described above, when Traffic_bust_1=100, Traffic_bust_1<T_th_1. Therefore, in this case, the prediction method determining unit 731 determines that low-computation-amount prediction is possible.

[0144] The small calculation amount prediction unit 733 predicts future traffic volume (Traffic_pred) based on the number of terminals and traffic volume, as in the following equation (10).

[0145] Traffic_pred=Traffic_old / Num_old×Num_new…(10)

[0146] Here, Traffic_old represents the past traffic volume, Num_old represents the past number of terminals, and Num_new represents the current number of terminals.

[0147] For example, in the case of flow 1 described in the flow information table shown in Fig. 15, the traffic volume in the first period is 200 kilobytes. Also, for flow 1 described in the flow information table shown in Fig. 15, the number of terminals in both the first period and the next period is 20, and the number of terminals remains unchanged. Therefore, in this case, the future traffic volume (Traffic_pred) is calculated as shown in the following formula (11).

[0148] Traffic_pred=200 / 20×20=200…(11)

[0149] In this way, in the above case, the small calculation amount prediction unit 733 predicts that the future (next period) traffic volume (Traffic_pred) of flow 1 will be 200 kilobytes.

[0150] The large-computational-amount prediction unit 732 and the small-computational-amount prediction unit 733 output the predicted future traffic volume values ​​for each flow to the total traffic calculation unit 740 (step S185). This completes the operation of the prediction execution unit 730 shown in the flowchart of FIG.

[0151] 13 , the total traffic calculation unit 740 acquires the values ​​of future traffic volume for each flow output from the large computational complexity prediction unit 732 and the small computational complexity prediction unit 733. The total traffic calculation unit 740 adds up the acquired values ​​of future traffic volume to obtain the value of the total traffic volume (hereinafter referred to as “total traffic volume”). The total traffic calculation unit 740 outputs the calculated value of the total traffic volume to the switching threshold determination unit 172.

[0152] The processing after the value of the total traffic volume is output from the total traffic calculation unit 740 of the prediction unit 171a to the switching threshold determination unit 172 (i.e., processing related to determining whether to perform route switching and processing related to controlling the implementation of route switching) is not described here because it is the same as the processing in the conventional communication system 1. In other words, the processing after the processing by the prediction unit 171a is the same as the processing related to the control of route switching by the switching threshold determination unit 172 and the switching determination unit 175 of the conventional communication system 1 described above with reference to Figures 9 to 12 and the like, and the processing related to route switching by the transfer device 15, and therefore not described here.

[0153] As described above, the switching instruction device 17a in the first embodiment of the present invention classifies each of the flows input from, for example, the lower-level device 3 during a predetermined period based on the characteristics of the traffic pattern (e.g., QoS) based on information defined in, for example, 5QI or S-NSSAI. For each classified flow, the switching instruction device 17a calculates the spread (burstiness) of traffic volume based on the time-series fluctuation (difference value) of the traffic volume. Based on the calculated spread (burstiness) of traffic volume, the switching instruction device 17a determines for each flow whether it is possible to predict future traffic volume using low-computational-amount prediction. As described above, low-computational-amount prediction is a prediction method with a relatively low computational amount, such as linear prediction.

[0154] The switching instruction device 17a performs low-computation-amount prediction for flows for which it is determined that future traffic volume can be predicted using low-computation-amount prediction. On the other hand, the switching instruction device 17a performs high-computation-amount prediction for flows for which it is determined that future traffic volume cannot be predicted using low-computation-amount prediction. As described above, high-computation-amount prediction is a prediction method that requires a relatively large amount of calculation, such as a prediction method using machine learning. The switching instruction device 17a sums up the future traffic volume calculated for each flow using low-computation-amount prediction or high-computation-amount prediction, and determines and controls the execution of route switching based on the total traffic volume.

[0155] As described above, the switching instruction device 17a according to the first embodiment of the present invention is configured to predict future traffic volume for each classified flow rather than predicting all flows collectively, thereby preventing degradation of communication quality. Furthermore, the switching instruction device 17a according to the first embodiment of the present invention is configured to perform low-computational-amount prediction, such as linear prediction, which requires less computational effort, for flows for which it is determined that future traffic volume can be predicted using low-computational-amount prediction, rather than performing high-computational-amount prediction using, for example, machine learning. By using such a configuration, the switching instruction device 17a can reduce the amount of computation required in predicting future traffic volume.

[0156] As a result, the switching instruction device 17a can reduce the amount of calculation required for predicting future traffic volume, thereby reducing communication delays that occur when prediction of future traffic volume and route switching are not completed before congestion occurs. In this way, the switching instruction device 17a in the first embodiment of the present invention can reduce communication delays while satisfying communication quality even when multiple flows are mixed.

[0157] <Second Embodiment> A communication system according to a second embodiment of the present invention will now be described. Note that the overall configuration of the communication system according to this embodiment is similar to the configuration of the conventional communication system 1 shown in the overall configuration diagram of Figure 1 described above, and therefore a description thereof will be omitted. Furthermore, the overall configuration of a mobile communication system in which the communication system according to this embodiment is applied to mobile communication is similar to the configuration of the conventional mobile communication system 10 shown in the overall configuration diagram of Figure 2 described above, and therefore a description thereof will be omitted.

[0158] Furthermore, the block diagram showing the functional configuration of the switching instruction device of the communication system in the second embodiment is the same as the block diagram showing the functional configuration of the switching instruction device 17 of the conventional communication system 1 shown in FIG. 3 described above. Hereinafter, the switching instruction device in the second embodiment will be referred to as switching instruction device 17b. The switching instruction device 17b is, for example, an information processing device such as a general-purpose computer. The switching instruction device 17b of this embodiment and the conventional switching instruction device 17 described above have different configurations of their prediction units. The configuration of the prediction unit 171b of the communication system in this embodiment will be described below.

[0159] 17 is a block diagram showing the functional configuration of a prediction unit 171b of a communication system according to the second embodiment of the present invention. As shown in Fig. 17, the prediction unit 171b includes a flow classification unit 710, a flow information storage unit 720, a prediction execution unit 730b, and a total traffic calculation unit 740.

[0160] As in the first embodiment described above, the flow classification unit 710 classifies flows input to the network 5 during a predetermined period based on traffic characteristics. The flow classification unit 710 classifies flows based on flow types based on information defined in, for example, 5QI or S-NSSAI. The flow classification unit 710 calculates the total number of terminals and traffic volume for each classified flow and stores them in a flow information table stored in the flow information storage unit 720.

[0161] The prediction execution unit 730b determines whether future traffic volume can be predicted using low-computation-amount prediction, which is a prediction method with a relatively low computational complexity, for each flow grouped (classified) by the flow classification unit 710. For flows for which it has determined that future traffic volume can be predicted using low-computation-amount prediction, the prediction execution unit 730b predicts future traffic volume using low-computation-amount prediction. Furthermore, for flows for which it has determined that future traffic volume cannot be predicted using low-computation-amount prediction, the prediction execution unit 730b predicts future traffic volume using high-computation-amount prediction, which is a prediction method with a relatively high computational complexity. The prediction execution unit 730b outputs the value of the future traffic volume predicted for each flow to the total traffic calculation unit 740.

[0162] In contrast to the conventional switching instruction device 17 that performs high-computational-amount prediction for all flows to be predicted, the switching instruction device 17b in this embodiment is configured to use low-computational-amount prediction for flows that can use low-computational-amount prediction, thereby enabling the switching instruction device 17b in this embodiment to reduce the amount of calculation required to predict future traffic volume.

[0163] As shown in FIG. 17, the prediction execution unit 730 b includes a prediction method determination unit 731 b, a large calculation amount prediction unit 732, and a small calculation amount prediction unit 733.

[0164] The prediction method determination unit 731b determines whether it is possible to predict future traffic volume using low-computational-amount prediction, based on the traffic volume and traffic pattern, for each flow grouped by the flow classification unit 710. More specifically, the prediction method determination unit 731b calculates the average traffic volume for each flow grouped by the flow classification unit 710, and determines whether it is possible to predict future traffic volume using low-computational-amount prediction, based on the calculation result.

[0165] As described above, the prediction execution unit 730b in this embodiment differs from the prediction execution unit 730 in the first embodiment in the configuration of the prediction method determination unit. The prediction method determination unit 731 in the first embodiment was configured to determine, for each flow, whether it is possible to predict future traffic volume using low-computational-complexity prediction based on time-series fluctuations in traffic volume (traffic volume spread (burstiness)). In contrast, the prediction method determination unit 731b in this embodiment is configured to determine, for each flow, whether it is possible to predict future traffic volume using low-computational-complexity prediction based on average traffic volume.

[0166] If the prediction method determination unit 731b determines that future traffic volume can be predicted using low-computational-amount prediction, it outputs information indicating the flow ID of the determined flow and the current traffic volume to the low-computational-amount prediction unit 733. On the other hand, if the prediction method determination unit 731b determines that future traffic volume cannot be predicted using low-computational-amount prediction, it outputs information indicating the flow ID of the determined flow and the current traffic volume to the high-computational-amount prediction unit 732.

[0167] The computationally intensive prediction unit 732 predicts future traffic volume using the computationally intensive prediction and outputs the value of the future traffic volume predicted for each flow to the total traffic calculation unit 740.

[0168] The small calculation amount prediction unit 733 predicts future traffic volume using the small calculation amount prediction, and outputs the value of the future traffic volume predicted for each flow to the total traffic calculation unit 740.

[0169] The process performed by the prediction execution unit 730b will be described in more detail below with examples. Fig. 18 is a flowchart showing the operation of the prediction execution unit 730b of the communication system according to the second embodiment of the present invention.

[0170] The prediction method determination unit 731b of the prediction execution unit 730b calculates the average traffic volume (Traffic_ave) in a predetermined period for each flow classified by the flow classification unit 710 (step S191).

[0171] The configuration of the communication system in this embodiment is not limited to a mobile network, but can also be applied to a wireless network, etc. In this case, the prediction method determination unit 731b may acquire information indicating the traffic volume from a Wi-Fi (registered trademark) access point, etc.

[0172] Next, the prediction method determination unit 731b determines whether the calculated average traffic volume is small. Specifically, the prediction method determination unit 731b determines whether the average traffic volume calculated in step S191 is smaller than a predetermined threshold (T_th) (step S192).

[0173] If it is determined that the average traffic volume is small (step S192, Yes), the prediction method determination unit 731b determines that it is possible to predict future traffic volume using low-computation-amount prediction. On the other hand, if it is determined that the average traffic volume is not small (step S192, No), the prediction method determination unit 731b determines that it is impossible to predict future traffic volume using low-computation-amount prediction.

[0174] A specific example will be described below. For example, it is assumed that the time-series fluctuation (traffic1_1, traffic1_2, . . . ) of the traffic volume (Traffic1) of flow 1 is as shown in the following equation (12).

[0175] Traffic1=[traffic1_1, traffic1_2,...]=[200,200,...]...(12)

[0176] At this time, the average traffic volume of flow 1 is calculated as in the following equation (13).

[0177] Traffic_ave_1=ave(traffic1_1+traffic1_2+...+traffic1_N)...(13)

[0178] Here, the value of N may be, for example, the maximum number of time-series data of traffic volume stored in the flow information storage unit 729, or may be any number (for example, N = 100). For example, when N = 10, the value of the average traffic volume (Traffic_ave_1) of flow 1 is expressed by the following formula (14).

[0179] Traffic_ave_1=ave(traffic1_1+traffic1_2+...+traffic1_10)=300...(14)

[0180] The prediction method determination unit 731b, for example, compares the calculated value of Traffic_ave_1 with a predetermined threshold (T_th) to determine whether the average traffic volume of flow 1 is low. The threshold T_th may be set to a predetermined fixed value (e.g., 800 kilobytes).

[0181] When the average traffic volume is low, it is possible to predict future traffic volume by low-computation-amount prediction using, for example, linear prediction, etc. Therefore, when the average traffic volume is low, the prediction method determination unit 731b causes the low-computation-amount prediction unit 733 to execute low-computation-amount prediction using, for example, linear prediction, etc. (step S193).

[0182] On the other hand, when the average traffic volume is not small, it is difficult to predict future traffic volume using low-computational-volume prediction, for example, linear prediction, etc. Therefore, when the average traffic volume is not small, the prediction method determination unit 731b causes the high-computational-volume prediction unit 732 to perform high-computational-volume prediction using, for example, machine learning (step S194).

[0183] When the value of Traffic_ave exceeds the threshold T_th as described above, the calculation-intensive prediction unit 732 performs calculation-intensive prediction (time-series prediction). Therefore, in this case, the calculation-intensive prediction unit 732 first calculates the prediction period. Here, the threshold T_th is set to 800 kilobytes.

[0184] For example, when Traffic_ave_1=300 as described above, the following equation (15) holds.

[0185] Traffic_ave_1<T_th_1...(15)

[0186] Therefore, in this case, the prediction method determining unit 731b determines to perform low-computation-amount prediction.

[0187] In this embodiment, the prediction method determination unit 731b is configured to determine the prediction method by comparing the average value of traffic volume in a predetermined period with a predetermined threshold, but the present invention is not limited to this configuration. For example, the prediction method determination unit 731b may be configured to determine the prediction method by comparing the maximum value of traffic volume in a predetermined period with a predetermined threshold.

[0188] The small calculation amount prediction unit 733 predicts future traffic volume (Traffic_pred) based on the number of terminals and traffic volume, as in the following equation (16).

[0189] Traffic_pred=Traffic_old / Num_old×Num_new…(16)

[0190] Here, Traffic_old represents the past traffic volume, Num_old represents the past number of terminals, and Num_new represents the current number of terminals.

[0191] For example, in the case of flow 1 described in the flow information table shown in Fig. 15, the traffic volume in the first period is 200 kilobytes. Also, for flow 1 described in the flow information table shown in Fig. 15, the number of terminals in both the first period and the next period is 20, and the number of terminals remains unchanged. Therefore, in this case, the future traffic volume (Traffic_pred) is calculated as shown in the following equation (17).

[0192] Traffic_pred=200 / 20×20=200…(17)

[0193] In this way, in the above case, the small calculation amount prediction unit 733 predicts that the future (next period) traffic volume (Traffic_pred) of flow 1 will be 200 kilobytes.

[0194] The large-computational-amount prediction unit 732 and the small-computational-amount prediction unit 733 output the predicted future traffic volume values ​​for each flow to the total traffic calculation unit 740 (step S195). This completes the operation of the prediction execution unit 730b shown in the flowchart of FIG.

[0195] 17 , the total traffic calculation unit 740 acquires the values ​​of future traffic volume for each flow output from the large computational complexity prediction unit 732 and the small computational complexity prediction unit 733. The total traffic calculation unit 740 adds up the acquired values ​​of future traffic volume to obtain the value of total traffic volume. The total traffic calculation unit 740 outputs the calculated value of total traffic volume to the switching threshold determination unit 172.

[0196] The processing after the value of the total traffic volume is output from the total traffic calculation unit 740 of the prediction unit 171b to the switching threshold determination unit 172 (i.e., processing related to determining whether to perform route switching and processing related to controlling the implementation of route switching) is not described here because it is the same as the processing in the conventional communication system 1. In other words, the subsequent processing after the processing by the prediction unit 171b is the same as the processing related to the control of route switching by the switching threshold determination unit 172 and the switching determination unit 175 of the conventional communication system 1 described with reference to the above-mentioned Figures 9 to 12, and the processing related to route switching by the transfer device 15, and therefore not described here.

[0197] As described above, the switching instruction device 17b in the second embodiment of the present invention classifies each of the flows input from, for example, the lower-level device 3 during a predetermined period based on the characteristics of the traffic pattern (e.g., QoS) based on information defined in, for example, 5QI or S-NSSAI. The switching instruction device 17b calculates the average traffic volume during the predetermined period for each classified flow. Based on the calculated average traffic volume, the switching instruction device 17b determines whether it is possible to predict future traffic volume for each flow using low-computational-amount prediction, which is a prediction method with a relatively low computational amount, such as linear prediction.

[0198] The switching instruction device 17b performs low-computation-amount prediction for flows for which it is determined that future traffic volume can be predicted using low-computation-amount prediction. On the other hand, for flows for which it is determined that future traffic volume cannot be predicted using low-computation-amount prediction, it performs high-computation-amount prediction, which is a prediction method with a relatively high computational load, such as a prediction method using machine learning, for example. The switching instruction device 17b adds up the future traffic volumes calculated for each flow using low-computation-amount prediction or high-computation-amount prediction, and determines and controls the execution of route switching based on the total traffic volume.

[0199] As described above, the switching instruction device 17b according to the second embodiment of the present invention is configured to predict future traffic volume for each classified flow rather than predicting all flows collectively, thereby preventing degradation of communication quality. Furthermore, the switching instruction device 17b according to the second embodiment of the present invention is configured to perform low-computational-amount prediction, such as linear prediction, which requires less computational overhead, for flows for which it is determined that future traffic volume can be predicted using low-computational-amount prediction, rather than performing high-computational-amount prediction using, for example, machine learning. By using such a configuration, the switching instruction device 17b can reduce the amount of computation required in predicting future traffic volume.

[0200] As a result, the switching instruction device 17b can reduce the amount of calculation required for predicting future traffic volume, thereby reducing communication delays that occur when prediction of future traffic volume and route switching are not completed before congestion occurs. In this way, the switching instruction device 17b in the second embodiment of the present invention can reduce communication delays while satisfying communication quality even when multiple flows are mixed.

[0201] <Third Embodiment> A communication system according to a third embodiment of the present invention will now be described. Note that the overall configuration of the communication system according to this embodiment is similar to the configuration of the conventional communication system 1 shown in the overall configuration diagram of Fig. 1 described above, and therefore a description thereof will be omitted. Furthermore, the overall configuration of a mobile communication system in which the communication system according to this embodiment is applied to mobile communication is also similar to the configuration of the conventional mobile communication system 10 shown in the overall configuration diagram of Fig. 2 described above, and therefore a description thereof will be omitted.

[0202] Furthermore, the block diagram showing the functional configuration of the switching instruction device of the communication system in the third embodiment is the same as the block diagram showing the functional configuration of the switching instruction device 17 of the conventional communication system 1 shown in FIG. 3 described above. Hereinafter, the switching instruction device in the third embodiment will be referred to as switching instruction device 17c. The switching instruction device 17c is, for example, an information processing device such as a general-purpose computer. The switching instruction device 17c of this embodiment and the conventional switching instruction device 17 described above have different configurations of their prediction units. The configuration of the prediction unit 171c of the communication system in this embodiment will be described below.

[0203] 19 is a block diagram showing the functional configuration of a prediction unit 171c of a communication system according to the third embodiment of the present invention. As shown in FIG. 19, the prediction unit 171c includes a flow classification unit 710, a flow information storage unit 720, a prediction execution unit 730c, and a total traffic calculation unit 740.

[0204] As in the first embodiment described above, the flow classification unit 710 classifies flows input to the network 5 during a predetermined period based on traffic characteristics. The flow classification unit 710 classifies flows by flow type based on information defined in, for example, 5QI or S-NSSAI. The flow classification unit 710 calculates the total number of terminals and traffic volume for each classified flow and stores them in a flow information table stored in the flow information storage unit 720.

[0205] The prediction execution unit 730c determines, for each flow grouped by the flow classification unit 710, whether it is possible to estimate future traffic volume using traffic estimation based on fluctuations in the number of terminals (hereinafter simply referred to as "traffic estimation"), which requires a relatively small amount of calculation. For flows for which it is determined that future traffic volume can be estimated using traffic estimation, the prediction execution unit 730c estimates future traffic volume using traffic estimation. Furthermore, for flows for which it is determined that future traffic volume cannot be estimated using traffic estimation, the prediction execution unit 730c predicts future traffic volume using high-computational-volume prediction, which is a prediction method requiring a relatively large amount of calculation. The prediction execution unit 730c outputs the value of the future traffic volume predicted (or estimated) for each flow to the total traffic calculation unit 740.

[0206] In contrast to the conventional switching instruction device 17 that performs a high-computational-amount prediction for all flows to be predicted, the switching instruction device 17c of this embodiment is configured to use traffic estimation for flows for which traffic estimation can be used, thereby enabling the switching instruction device 17c of this embodiment to reduce the amount of calculation required for predicting future traffic volume.

[0207] As shown in FIG. 19, the prediction execution unit 730 c includes a prediction method determination unit 731 c, a large calculation amount prediction unit 732 , and a traffic estimation unit 734 .

[0208] The prediction method determination unit 731c determines whether it is possible to estimate future traffic volume using traffic estimation for each flow grouped (classified) by the flow classification unit 710. More specifically, the prediction method determination unit 731c calculates the changed number of terminals for each flow grouped by the flow classification unit 710, and determines whether it is possible to estimate future traffic volume using traffic estimation based on the calculation result.

[0209] As described above, the prediction execution unit 730c in this embodiment differs from the prediction execution unit 730 in the first embodiment and the prediction execution unit 730b in the second embodiment in the configuration of the prediction method determination unit. The prediction method determination unit 731 in the first embodiment was configured to determine, for each flow, whether it is possible to predict future traffic volume using low-computation-complexity prediction based on time-series fluctuations in traffic volume (traffic volume spread (burstiness)). The prediction method determination unit 731b in the second embodiment was configured to determine, for each flow, whether it is possible to predict future traffic volume using low-computation-complexity prediction based on average traffic volume. In contrast, the prediction method determination unit 731c in this embodiment is configured to determine, for each flow, whether it is possible to estimate future traffic volume using traffic estimation based on a fluctuating number of terminals.

[0210] If the changed number of terminals is 0 (i.e., if the number of terminals has not changed), the prediction method determination unit 731c determines that the future traffic volume can be estimated using traffic estimation, and outputs information indicating the flow ID of the determined flow and the current traffic volume to the traffic estimation unit 734. On the other hand, if the changed number of terminals is 1 or more (i.e., if the number of terminals has changed), the prediction method determination unit 731b determines that the future traffic volume cannot be estimated using traffic estimation, and outputs information indicating the flow ID of the determined flow and the current traffic volume to the high-computational-volume prediction unit 732.

[0211] The computationally intensive prediction unit 732 predicts future traffic volume using the computationally intensive prediction and outputs the value of the future traffic volume predicted for each flow to the total traffic calculation unit 740.

[0212] The traffic estimation unit 734 estimates future traffic volume using the traffic estimation, and outputs the value of the future traffic volume estimated for each flow to the total traffic calculation unit 740.

[0213] The process performed by the prediction execution unit 730c will be described in more detail below with examples. Fig. 20 is a flowchart showing the operation of the prediction execution unit 730c of the communication system according to the third embodiment of the present invention.

[0214] The prediction method determination unit 731c of the prediction execution unit 730c calculates the number of terminals (Num_val) that has changed in a predetermined period for each flow classified by the flow classification unit 710 (step S201).

[0215] Next, the prediction method determination unit 731c determines whether there has been a change in the number of terminals. Specifically, the prediction method determination unit 731c determines whether the changed number of terminals (Num_val) calculated in step S201 above is 0 (step S202).

[0216] If it is determined that there is no change in the number of terminals (step S202, Yes), the prediction method determination unit 731c determines that it is possible to estimate future traffic volume using traffic estimation. On the other hand, if it is determined that there is a change in the number of terminals (step S202, No), the prediction method determination unit 731c determines that it is impossible to estimate future traffic volume using traffic estimation.

[0217] A more specific example will be described below. For example, the prediction method determination unit 731c checks whether or not there is a change in the number of terminals for an arbitrary period (R_n). Note that the value of R_n needs to be sufficiently large to include the traffic cycle.

[0218] For example, consider the fluctuation in the number of terminals over a 10-period period for the traffic of flow 2 recorded in the flow information table shown in Fig. 15. The maximum value of the fluctuating number of terminals (Num_val_2) for flow 2 over the 10-period period is 0, as shown in the following equation (18). This is because, as shown in Fig. 15, the number of terminals for flow 2 has remained unchanged at 10 throughout the 10-period period.

[0219] Num_val_2=max{(num2_2-num2_1)+(num2_3-num2_2)+...+(num2_11-num2_10)}=0...(18)

[0220] In this way, for a flow in which the number of terminals does not change (Num_val = 0), it is possible to estimate future traffic volume by, for example, linear prediction, etc. Therefore, when there is no change in the number of terminals, the prediction method determination unit 731c causes the traffic estimation unit 734 to execute processing to estimate future traffic volume by, for example, linear prediction, etc. (step S203).

[0221] On the other hand, for a flow in which the number of terminals fluctuates (Num_val>0), it is difficult to estimate future traffic volume by, for example, linear prediction, etc. Therefore, when the number of terminals fluctuates, the prediction method determination unit 731c causes the high-computation-amount prediction unit 732 to perform high-computation-amount prediction using, for example, machine learning (step S204).

[0222] The large-computational-amount prediction unit 732 and the traffic estimation unit 734 output the future traffic volume values ​​for each flow, which are the results of the prediction (or estimation), to the total traffic calculation unit 740 (step S205). This completes the operation of the prediction execution unit 730c shown in the flowchart of FIG.

[0223] 19 , the total traffic calculation unit 740 acquires the values ​​of future traffic volume for each flow output from the high-computational-volume prediction unit 732 and the traffic estimation unit 734. The total traffic calculation unit 740 adds up the acquired values ​​of future traffic volume to obtain the value of total traffic volume. The total traffic calculation unit 740 outputs the calculated value of total traffic volume to the switching threshold determination unit 172.

[0224] The processing after the value of the total traffic volume is output from the total traffic calculation unit 740 of the prediction unit 171c to the switching threshold determination unit 172 (i.e., processing related to determining whether to perform route switching and processing related to controlling the implementation of route switching) is not described here because it is the same as the processing in the conventional communication system 1. In other words, the subsequent processing after the processing by the prediction unit 171c is the same as the processing related to the control of route switching by the switching threshold determination unit 172 and the switching determination unit 175 of the conventional communication system 1 described above with reference to Figures 9 to 12 and the like, and the processing related to route switching by the transfer device 15, and therefore not described here.

[0225] As described above, the switching instruction device 17c in the third embodiment of the present invention classifies each of the flows input from, for example, the lower-level device 3 during a predetermined period based on the characteristics of the traffic pattern (e.g., QoS) based on information defined in, for example, 5QI or S-NSSAI. The switching instruction device 17c calculates the number of terminals that has changed during the predetermined period for each classified flow. Based on the changed number of terminals, the switching instruction device 17c determines for each flow whether it is possible to estimate future traffic volume using traffic estimation such as linear prediction.

[0226] The switching instruction device 17c performs traffic estimation for flows for which it is determined that future traffic volume can be estimated using traffic estimation. On the other hand, for flows for which it is determined that future traffic volume cannot be estimated using traffic estimation, the switching instruction device 17c performs high-computational-volume prediction, which is a prediction method with a relatively high computational volume, such as a prediction method using machine learning. The switching instruction device 17c adds up the future traffic volumes calculated for each flow by traffic estimation or high-computational-volume prediction, and determines and controls the execution of route switching based on the total traffic volume.

[0227] As described above, the switching instruction device 17c according to the third embodiment of the present invention is configured to predict future traffic volume for each classified flow rather than predicting all flows collectively, thereby preventing degradation of communication quality. Furthermore, the switching instruction device 17c according to the third embodiment of the present invention is configured to perform traffic estimation such as linear prediction, which requires less computational effort, for flows for which it is determined that future traffic volume can be estimated using traffic estimation, rather than performing computationally intensive prediction using, for example, machine learning. By using such a configuration, the switching instruction device 17c can reduce the computational effort required for predicting future traffic volume.

[0228] As a result, the switching instruction device 17c can reduce the amount of calculation required for predicting future traffic volume, thereby reducing communication delays that occur when prediction of future traffic volume and route switching are not completed before congestion occurs. In this way, the switching instruction device 17c in the third embodiment of the present invention can reduce communication delays while satisfying communication quality even when multiple flows are mixed.

[0229] <Fourth embodiment> A communication system according to a fourth embodiment of the present invention will now be described. Note that the overall configuration of the communication system according to this embodiment is similar to the configuration of the conventional communication system 1 shown in the overall configuration diagram of Fig. 1 described above, and therefore a description thereof will be omitted. Furthermore, the overall configuration of a mobile communication system in which the communication system according to this embodiment is applied to mobile communication is also similar to the configuration of the conventional mobile communication system 10 shown in the overall configuration diagram of Fig. 2 described above, and therefore a description thereof will be omitted.

[0230] Furthermore, the block diagram showing the functional configuration of the switching instruction device of the communication system in the fourth embodiment is the same as the block diagram showing the functional configuration of the switching instruction device 17 of the conventional communication system 1 shown in FIG. 3 described above. Hereinafter, the switching instruction device in the fourth embodiment will be referred to as switching instruction device 17d. The switching instruction device 17d is, for example, an information processing device such as a general-purpose computer. The switching instruction device 17d of this embodiment and the conventional switching instruction device 17 described above have different configurations of their prediction units. The configuration of the prediction unit 171d of the communication system in this embodiment will be described below.

[0231] 21 is a block diagram showing the functional configuration of a prediction unit 171d in a communication system according to the fourth embodiment of the present invention. As shown in Fig. 21, the prediction unit 171d includes a flow classification unit 710, a flow information storage unit 720, a prediction execution unit 730d, and a total traffic calculation unit 740.

[0232] As in the first embodiment described above, the flow classification unit 710 classifies flows input to the network 5 during a predetermined period based on traffic characteristics. The flow classification unit 710 classifies flows by flow type based on information defined in, for example, 5QI or S-NSSAI. The flow classification unit 710 calculates the total number of terminals and traffic volume for each classified flow and stores them in a flow information table stored in the flow information storage unit 720.

[0233] The prediction execution unit 730d determines, for each flow grouped by the flow classification unit 710, whether future traffic volume can be estimated using traffic estimation, which requires a relatively small amount of calculation. For flows for which it has determined that future traffic volume can be estimated using traffic estimation, the prediction execution unit 730d estimates future traffic volume using traffic estimation. Furthermore, for flows for which it has determined that future traffic volume cannot be estimated using traffic estimation, the prediction execution unit 730d predicts future traffic volume using high-computational-volume prediction, which is a prediction method requiring a relatively large amount of calculation. The prediction execution unit 730d outputs the value of the future traffic volume predicted (or estimated) for each flow to the total traffic calculation unit 740.

[0234] In contrast to the conventional switching instruction device 17 that performs a high-computational-amount prediction for all flows to be predicted, the switching instruction device 17d of this embodiment is configured to use traffic estimation for flows for which traffic estimation can be used, similar to the switching instruction device 17c of the third embodiment described above, thereby enabling the switching instruction device 17d of this embodiment to reduce the amount of calculation required for predicting future traffic volume.

[0235] 21 , the prediction execution unit 730d includes a traffic period calculation unit 735, a prediction method determination unit 731d, a high-computational-volume prediction unit 732, and a traffic estimation unit 734d. As described above, the prediction execution unit 730d in this embodiment is configured such that the traffic period calculation unit 735 is added to the prediction execution unit 730c in the third embodiment.

[0236] The traffic period calculation unit 735 calculates the traffic period of each flow and outputs the calculated traffic period value of each flow to the prediction method determination unit 731d.

[0237] The traffic period calculation unit 735 identifies the traffic pattern of each flow based on, for example, 5QI or S-NSSAI. The traffic period calculation unit 735 calculates the traffic period for each flow based on the identified traffic pattern. Note that the traffic period calculation unit 735 may also calculate the traffic period for each flow based on the results of spectral analysis of time-series traffic.

[0238] The prediction method determination unit 731d acquires the value of the traffic period of each flow output from the traffic period calculation unit 735. The prediction method determination unit 731d determines whether it is possible to estimate future traffic volume using traffic estimation for each flow grouped (classified) by the flow classification unit 710. More specifically, the prediction method determination unit 731d calculates the changed number of terminals for each flow grouped by the flow classification unit 710, and determines whether it is possible to estimate future traffic volume using traffic estimation based on the calculation result.

[0239] At this time, the prediction method determination unit 731d in this embodiment calculates the number of terminals that has changed for each traffic cycle obtained from the traffic cycle calculation unit 735. As a result, for a flow in which the number of terminals does not change, the traffic estimation unit 734d (described later) can perform estimation using an algorithm that estimates future traffic volume based on fluctuations in traffic volume accompanying traffic cycles. As a result, the prediction method determination unit 731d can perform traffic estimation using traffic cycles, thereby narrowing the range in which fluctuations in traffic volume are checked (i.e., reducing the amount of input data), and thereby further reducing the amount of calculation in the process of predicting future traffic volume.

[0240] Furthermore, like the prediction execution unit 730c in the third embodiment, the prediction execution unit 730d in this embodiment differs from the prediction execution unit 730 in the first embodiment and the prediction execution unit 730b in the second embodiment in the configuration of the prediction method determination unit. The prediction method determination unit 731 in the first embodiment was configured to determine, for each flow, whether it is possible to predict future traffic volume using low-computational complexity prediction based on time-series fluctuations in traffic volume (traffic volume spread (burstiness)). The prediction method determination unit 731b in the second embodiment was configured to determine, for each flow, whether it is possible to predict future traffic volume using low-computational complexity prediction based on average traffic volume. In contrast, the prediction method determination unit 731d in this embodiment is configured to determine, for each flow, whether it is possible to estimate future traffic volume using traffic estimation.

[0241] If the number of terminals that has changed within the traffic cycle is zero (i.e., the number of terminals has not changed within the traffic cycle), the prediction method determination unit 731c determines that it is possible to estimate future traffic volume using traffic estimation, and outputs information indicating the flow ID of the determined flow and the current traffic volume to the traffic estimation unit 734d. On the other hand, if the number of terminals that has changed within the traffic cycle is one or more (i.e., the number of terminals has changed within the traffic cycle), the prediction method determination unit 731b determines that it is impossible to estimate future traffic volume using traffic estimation, and outputs information indicating the flow ID of the determined flow and the current traffic volume to the high-computational-complexity prediction unit 732.

[0242] The computationally intensive prediction unit 732 predicts future traffic volume using the computationally intensive prediction and outputs the value of the future traffic volume predicted for each flow to the total traffic calculation unit 740.

[0243] The traffic estimation unit 734d estimates future traffic volume using traffic estimation. Here, the traffic estimation unit 734d performs traffic estimation using an algorithm that estimates future traffic volume based on fluctuations in traffic volume associated with traffic cycles. The traffic estimation unit 734d outputs the value of the future traffic volume estimated for each flow to the total traffic calculation unit 740.

[0244] The process performed by the prediction execution unit 730d will be described in more detail below with examples. Fig. 22 is a flowchart showing the operation of the prediction execution unit 730d in the communication system according to the fourth embodiment of the present invention.

[0245] The traffic period calculation unit 735 of the prediction execution unit 730d calculates the traffic period of each flow. The traffic period calculation unit 735 outputs the calculated traffic period value of each flow to the prediction method determination unit 731d. The prediction method determination unit 731d acquires the traffic period value of each flow output from the traffic period calculation unit 735. The prediction method determination unit 731d calculates the number of terminals (Num_val) that has changed within the traffic period for each flow classified by the flow classification unit 710 (step S211).

[0246] Next, the prediction method determination unit 731d determines whether the number of terminals has changed within the traffic cycle. Specifically, the prediction method determination unit 731d determines whether the changed number of terminals (Num_val) calculated in step S211 is 0 (step S212).

[0247] If it is determined that there is no change in the number of terminals within the traffic cycle (step S212, Yes), the prediction method determination unit 731d determines that it is possible to estimate future traffic volume using traffic estimation.On the other hand, if it is determined that there is a change in the number of terminals within the traffic cycle (step S212, No), the prediction method determination unit 731c determines that it is impossible to estimate future traffic volume using traffic estimation.

[0248] A more specific example will be described below. For example, assume that the traffic period (Traffic_period) of flow 2 recorded in the flow information table shown in Fig. 15 is 5 periods. In this case, the maximum value of the number of terminals (Num_val_2) that has changed in flow 2 over the 5 periods is 0, as shown in the following equation (19). This is because, as shown in Fig. 15, the number of terminals in flow 2 has remained unchanged at 10 throughout the 5 periods.

[0249] Num_val_2=max{(num2_2-num2_1)+(num2_3-num2_2)+...+(num2_6-num2_5)}=0...(19)

[0250] In this way, for a flow in which the number of terminals does not change (Num_val = 0), the prediction method determination unit 731d causes the traffic estimation unit 734d to perform traffic estimation in which, for example, the traffic volume one period ago (i.e., the traffic volume five periods ago) is used as an estimated value of the future traffic volume (step S213). That is, the estimated future traffic volume (T_pred) is expressed by the following equation (20).

[0251] T_pred=Traffic_2_0=Traffic_2_5=0...(20)

[0252] On the other hand, for flows in which the number of terminals fluctuates (Num_val>0), it is difficult to estimate future traffic volume by, for example, linear prediction, etc. Therefore, when the number of terminals fluctuates, the prediction method determination unit 731d causes the high-computation-amount prediction unit 732 to perform high-computation-amount prediction using machine learning (step S214).

[0253] The large-computational-amount prediction unit 732 and the traffic estimation unit 734d output the future traffic volume values ​​for each flow, which are the results of the prediction (or estimation), to the total traffic calculation unit 740 (step S215). This completes the operation of the prediction execution unit 730d shown in the flowchart of FIG.

[0254] 21 , the total traffic calculation unit 740 acquires the values ​​of future traffic volume for each flow output from the high-computational-volume prediction unit 732 and the traffic estimation unit 734d. The total traffic calculation unit 740 adds up the acquired values ​​of future traffic volume to obtain the value of total traffic volume. The total traffic calculation unit 740 outputs the calculated value of total traffic volume to the switching threshold determination unit 172.

[0255] The processing after the value of the total traffic volume is output from the total traffic calculation unit 740 of the prediction unit 171d to the switching threshold determination unit 172 (i.e., processing related to determining whether to perform route switching and processing related to controlling the implementation of route switching) is not described here because it is the same as the processing in the conventional communication system 1. In other words, the subsequent processing after the processing by the prediction unit 171d is the same as the processing related to the control of route switching by the switching threshold determination unit 172 and the switching determination unit 175 of the conventional communication system 1 described with reference to the above-mentioned Figures 9 to 12, and the processing related to route switching by the transfer device 15, and therefore not described here.

[0256] As described above, the switching instruction device 17d in the fourth embodiment of the present invention classifies each of the flows input from, for example, the lower-level device 3 during a predetermined period based on the characteristics of the traffic pattern (e.g., QoS) based on information defined in, for example, 5QI or S-NSSAI. The switching instruction device 17d calculates the number of terminals that has changed within the predetermined period for each classified flow. Based on the changed number of terminals, the switching instruction device 17d determines for each flow whether it is possible to estimate future traffic volume using traffic estimation.

[0257] At this time, the prediction method determination unit 731d in this embodiment calculates the changed number of terminals for each traffic cycle obtained from the traffic cycle calculation unit 735. The prediction method determination unit 731c can perform traffic estimation using the traffic cycle, thereby reducing the amount of input data and further reducing the amount of calculation in the process of predicting future traffic volume.

[0258] The switching instruction device 17d performs traffic estimation for flows for which it is determined that future traffic volume can be estimated using traffic estimation. On the other hand, for flows for which it is determined that future traffic volume cannot be estimated using traffic estimation, the switching instruction device 17d performs high-computational-volume prediction, which is a prediction method with a relatively high computational volume, such as a prediction method using machine learning. The switching instruction device 17d adds up the future traffic volumes calculated for each flow by traffic estimation or high-computational-volume prediction, and determines and controls the execution of route switching based on the total traffic volume.

[0259] As described above, the switching instruction device 17d according to the fourth embodiment of the present invention is configured to predict future traffic volume for each classified flow rather than predicting all flows collectively, thereby preventing degradation of communication quality. Furthermore, the switching instruction device 17d according to the fourth embodiment of the present invention is configured to perform traffic estimation such as linear prediction, which requires less computational effort, for flows for which it is determined that future traffic volume can be estimated using traffic estimation, rather than performing computationally intensive prediction using, for example, machine learning. By using such a configuration, the switching instruction device 17d can reduce the amount of computation required in predicting future traffic volume.

[0260] As a result, the switching instruction device 17d can reduce the amount of calculation required for predicting future traffic volume, thereby reducing communication delays that occur when prediction of future traffic volume and route switching are not completed before congestion occurs. In this way, the switching instruction device 17d in the fourth embodiment of the present invention can reduce communication delays while satisfying communication quality even when multiple flows are mixed.

[0261] <Fifth Embodiment> A communication system according to a fifth embodiment of the present invention will now be described. Note that the overall configuration of the communication system according to this embodiment is similar to the configuration of the conventional communication system 1 shown in the overall configuration diagram of Fig. 1 described above, and therefore a description thereof will be omitted. Furthermore, the overall configuration of a mobile communication system in which the communication system according to this embodiment is applied to mobile communication is also similar to the configuration of the conventional mobile communication system 10 shown in the overall configuration diagram of Fig. 2 described above, and therefore a description thereof will be omitted.

[0262] Furthermore, the block diagram showing the functional configuration of the switching instruction device of the communication system in the fifth embodiment is the same as the block diagram showing the functional configuration of the switching instruction device 17 of the conventional communication system 1 shown in FIG. 3 described above. Hereinafter, the switching instruction device in the fifth embodiment will be referred to as switching instruction device 17e. The switching instruction device 17e is, for example, an information processing device such as a general-purpose computer. The switching instruction device 17e of this embodiment and the conventional switching instruction device 17 described above have different configurations of their prediction units. The configuration of the prediction unit 17e of the communication system in this embodiment will be described below.

[0263] 23 is a block diagram showing the functional configuration of a prediction unit 171e of a communication system according to the fifth embodiment of the present invention. As shown in Fig. 23, the prediction unit 171e includes a flow classification unit 710, a flow information storage unit 720, a prediction execution unit 730e, and a total traffic calculation unit 740.

[0264] As in the first embodiment described above, the flow classification unit 710 classifies flows input to the network 5 during a predetermined period based on traffic characteristics. The flow classification unit 710 classifies flows by flow type based on information defined in, for example, 5QI or S-NSSAI. The flow classification unit 710 calculates the total number of terminals and traffic volume for each classified flow and stores them in a flow information table stored in the flow information storage unit 720.

[0265] The prediction execution unit 730e calculates a traffic period for each flow grouped by the flow classification unit 710, and predicts future traffic volume for each traffic period using high-computational-volume prediction. The prediction execution unit 730e outputs the value of the future traffic volume predicted for each flow to the total traffic calculation unit 740.

[0266] The switching instruction device 17e in this embodiment sets the traffic period equal to the input data, thereby reducing the number of input data used in the prediction process of future traffic volume, and therefore reducing the amount of calculation in the prediction process.

[0267] 23, the prediction execution unit 730d includes a traffic period calculation unit 735 and a high-computational-volume prediction unit 732e. As described above, the prediction execution unit 730e in this embodiment is configured such that the prediction method determination unit 731d and the traffic estimation unit 734d are omitted from the prediction execution unit 730e in the fourth embodiment.

[0268] The traffic period calculation unit 735 calculates the traffic period of each flow and outputs the calculated traffic period value of each flow to the high-computational-amount prediction unit 732e.

[0269] The traffic period calculation unit 735 identifies the traffic pattern of each flow based on, for example, 5QI or S-NSSAI. The traffic period calculation unit 735 calculates the traffic period for each flow based on the identified traffic pattern. Note that the traffic period calculation unit 735 may also calculate the traffic period for each flow based on the results of spectral analysis of time-series traffic.

[0270] The high-computational-amount prediction unit 732e acquires the value of the traffic period of each flow output from the traffic period calculation unit 735. The high-computational-amount prediction unit 732 predicts future traffic volume for each traffic period using the high-computational-amount prediction. The high-computational-amount prediction unit 732 outputs the value of future traffic volume predicted for each flow to the total traffic calculation unit 740.

[0271] The processing performed by the prediction execution unit 730e will be described below. Fig. 24 is a flowchart showing the operation of the prediction execution unit 730e of the communication system according to the fifth embodiment of the present invention.

[0272] The traffic period calculation unit 735 of the prediction execution unit 730e calculates the traffic period of each flow (step S221). The traffic period calculation unit 735 outputs the calculated traffic period value of each flow to the high-computational-amount prediction unit 732e. The high-computational-amount prediction unit 732e acquires the traffic period value of each flow output from the traffic period calculation unit 735. The high-computational-amount prediction unit 732 predicts future traffic volume for each traffic period using the high-computational-amount prediction (step S222). The high-computational-amount prediction unit 732 outputs the value of future traffic volume predicted for each flow to the total traffic calculation unit 740 (step S223).

[0273] This completes the operation of the prediction execution unit 730e shown in the flowchart of FIG.

[0274] 23 , the total traffic calculation unit 740 acquires the value of future traffic volume for each flow output from the high-computational-volume prediction unit 732. The total traffic calculation unit 740 adds up the acquired values ​​of future traffic volume to obtain the value of total traffic volume. The total traffic calculation unit 740 outputs the calculated value of total traffic volume to the switching threshold determination unit 172.

[0275] The processing after the value of the total traffic volume is output from the total traffic calculation unit 740 of the prediction unit 171e to the switching threshold determination unit 172 (i.e., processing related to determining whether to perform route switching and processing related to controlling the implementation of route switching) is not described here because it is the same as the processing in the conventional communication system 1. In other words, the subsequent processing after the processing by the prediction unit 171e is the same as the processing related to the control of route switching by the switching threshold determination unit 172 and the switching determination unit 175 of the conventional communication system 1 described with reference to the above-mentioned Figures 9 to 12, and the processing related to route switching by the transfer device 15, and therefore not described here.

[0276] As described above, the switching instruction device 17e in the fifth embodiment of the present invention classifies each of the flows input from, for example, the lower-level device 3 during a predetermined period based on the characteristics of the traffic pattern (e.g., QoS) based on information defined in, for example, 5QI or S-NSSAI, etc. The switching instruction device 17d performs high-computational-load prediction, which is a prediction method with a relatively large amount of calculation, such as a prediction method using machine learning, for each classified flow.

[0277] At this time, the prediction execution unit 730e calculates a traffic period for each flow grouped by the flow classification unit 710, and predicts future traffic volume for each traffic period using a computationally intensive prediction. By setting the traffic period equal to the input data, the switching instruction device 17e can reduce the amount of input data used in the prediction process for future traffic volume, thereby reducing the amount of calculation in the prediction process.

[0278] The switching instruction device 17e sums up the future traffic volumes calculated for each flow by the high-computation-volume prediction, and determines whether to perform route switching and controls its execution based on the total traffic volume.

[0279] As described above, the switching instruction device 17e according to the fifth embodiment of the present invention is configured to perform predictions for each classified flow rather than predicting all flows collectively when predicting future traffic volume, thereby preventing a deterioration in communication quality. Furthermore, the switching instruction device 17e according to the fifth embodiment of the present invention is configured to perform computationally intensive predictions for each traffic cycle using, for example, machine learning. By using such a configuration, the switching instruction device 17d can reduce the amount of input data, thereby reducing the amount of calculation required in the process of predicting future traffic volume.

[0280] As a result, the switching instruction device 17e can reduce the amount of calculation required for predicting future traffic volume, thereby reducing communication delays that occur when prediction of future traffic volume and route switching are not completed before congestion occurs. In this way, the switching instruction device 17e in the fifth embodiment of the present invention can reduce communication delays while satisfying communication quality even when multiple flows are mixed.

[0281] <Sixth Embodiment> A communication system according to a sixth embodiment of the present invention will now be described. Note that the overall configuration of the communication system according to this embodiment is similar to the configuration of the conventional communication system 1 shown in the overall configuration diagram of Fig. 1 described above, and therefore a description thereof will be omitted. Furthermore, the overall configuration of a mobile communication system in which the communication system according to this embodiment is applied to mobile communication is also similar to the configuration of the conventional mobile communication system 10 shown in the overall configuration diagram of Fig. 2 described above, and therefore a description thereof will be omitted.

[0282] Furthermore, the block diagram showing the functional configuration of the switching instruction device of the communication system in the sixth embodiment is the same as the block diagram showing the functional configuration of the switching instruction device 17 of the conventional communication system 1 shown in FIG. 3 described above. Hereinafter, the switching instruction device in the sixth embodiment will be referred to as switching instruction device 17f. The switching instruction device 17f is, for example, an information processing device such as a general-purpose computer. The switching instruction device 17f of this embodiment and the conventional switching instruction device 17 described above have different configurations of their prediction units. The configuration of the prediction unit 171f of the communication system in this embodiment will be described below.

[0283] 25 is a block diagram showing the functional configuration of a prediction unit 171f of a communication system according to the sixth embodiment of the present invention. As shown in Fig. 25, the prediction unit 171f includes a flow classification unit 710, a flow information storage unit 720, a prediction execution unit 730f, and a total traffic calculation unit 740.

[0284] As in the first embodiment described above, the flow classification unit 710 classifies flows input to the network 5 during a predetermined period based on traffic characteristics. The flow classification unit 710 classifies flows by flow type based on information defined in, for example, 5QI or S-NSSAI. The flow classification unit 710 calculates the total number of terminals and traffic volume for each classified flow and stores them in a flow information table stored in the flow information storage unit 720.

[0285] The prediction execution unit 730f determines whether future traffic volume can be predicted using low-computational-amount prediction, which is a prediction method with a relatively low computational-amount of computation, for each flow grouped (classified) by the flow classification unit 710. For flows for which it has determined that future traffic volume can be predicted using low-computational-amount prediction, the prediction execution unit 730f predicts future traffic volume using low-computational-amount prediction. Furthermore, for flows for which it has determined that future traffic volume cannot be predicted using low-computational-amount prediction, the prediction execution unit 730f further determines whether future traffic volume can be estimated using traffic estimation.

[0286] For flows for which it is determined that future traffic volume can be estimated using traffic estimation, the prediction execution unit 730f estimates future traffic volume using traffic estimation. For flows for which it is determined that future traffic volume cannot be estimated using traffic estimation, the prediction execution unit 730f predicts future traffic volume using high-computational-volume prediction, which is a prediction method with a relatively high computational load. The prediction execution unit 730f outputs the value of the future traffic volume predicted (or estimated) for each flow to the total traffic calculation unit 740.

[0287] In contrast to the conventional switching instruction device 17 that performs high-computational-amount prediction for all flows to be predicted, the switching instruction device 17f of this embodiment is configured to use low-computational-amount prediction for flows for which low-computational-amount prediction can be used, and to use traffic estimation for flows for which traffic estimation can be used. As a result, the switching instruction device 17f of this embodiment can reduce the amount of calculation required to predict future traffic volume.

[0288] As shown in FIG. 25, the prediction execution unit 730f is composed of a first prediction method determination unit 731f, a traffic period calculation unit 735f, a second prediction method determination unit 736, a high calculation amount prediction unit 732f, a low calculation amount prediction unit 733, and a traffic estimation unit 734f.

[0289] The configuration of the prediction execution unit 730f in this embodiment is a combination of the configuration of the prediction execution unit 730 in the first embodiment shown in Figure 13 above and the configuration of the prediction execution unit 730d in the fourth embodiment shown in Figure 21 above.

[0290] It is also possible to use a configuration that combines, for example, the configuration of the prediction execution unit 730b in the second embodiment shown in FIG. 17 and the configuration of the prediction execution unit 730d in the fourth embodiment shown in FIG. 21, or a configuration that combines the configuration of the prediction execution unit 730 in the first embodiment shown in FIG. 13 and the configuration of the prediction execution unit 730c in the third embodiment shown in FIG. 19.

[0291] The first prediction method determination unit 731f determines whether or not it is possible to predict future traffic volume using low-computational-amount prediction, based on the traffic volume and traffic pattern, for each flow grouped by the flow classification unit 710. More specifically, the first prediction method determination unit 731f calculates the time-series fluctuation of traffic volume (traffic volume spread (burstiness)) for each flow grouped by the flow classification unit 710, and determines whether or not it is possible to predict future traffic volume using low-computational-amount prediction, based on the calculation result.

[0292] If the first prediction method determination unit 731f determines that future traffic volume can be predicted using low-computation-amount prediction, it outputs information indicating the flow ID of the determined flow and the current traffic volume to the low-computation-amount prediction unit 733. On the other hand, if the prediction method determination unit 731f determines that future traffic volume cannot be predicted using low-computation-amount prediction, it outputs information indicating the flow ID of the determined flow and the current traffic volume to the traffic period calculation unit 735f.

[0293] The traffic period calculation unit 735 acquires information indicating the current traffic volume output from the first prediction method determination unit 731f. The traffic period calculation unit 735 calculates the traffic period of each flow. The traffic period calculation unit 735 outputs the calculated traffic period value of each flow and information indicating the current traffic volume to the second prediction method determination unit 736.

[0294] The traffic period calculation unit 735f identifies the traffic pattern of each flow based on, for example, 5QI or S-NSSAI. The traffic period calculation unit 735f calculates the traffic period for each flow based on the identified traffic pattern. Note that the traffic period calculation unit 735f may also calculate the traffic period for each flow based on the results of spectral analysis of time-series traffic.

[0295] The second prediction method determination unit 736 acquires information indicating the traffic period value and current traffic volume of each flow output from the traffic period calculation unit 735f. The second prediction method determination unit 736 determines whether it is possible to estimate future traffic volume using traffic estimation for each flow grouped (classified) by the flow classification unit 710. More specifically, the second prediction method determination unit 736 calculates the changed number of terminals for each flow grouped by the flow classification unit 710, and determines whether it is possible to estimate future traffic volume using traffic estimation based on the calculation result.

[0296] At this time, the second prediction method determination unit 736 in this embodiment calculates the number of terminals that has changed for each traffic period obtained from the traffic period calculation unit 735f. As a result, for flows in which the number of terminals does not change, the traffic estimation unit 734f (described later) can estimate future traffic volume using an algorithm that estimates future traffic volume based on fluctuations in traffic volume associated with traffic periods. As a result, the second prediction method determination unit 736 can perform traffic estimation using traffic periods, thereby narrowing the range in which fluctuations in traffic volume are checked (i.e., reducing the amount of input data), thereby further reducing the amount of calculation in the process of predicting future traffic volume.

[0297] If the number of terminals that changed within the traffic cycle is zero (i.e., the number of terminals has not changed within the traffic cycle), the second prediction method determination unit 736 determines that it is possible to estimate future traffic volume using traffic estimation, and outputs information indicating the flow ID of the determined flow and the current traffic volume to the traffic estimation unit 734f.On the other hand, if the number of terminals that changed within the traffic cycle is one or more (i.e., the number of terminals has changed within the traffic cycle), the second prediction method determination unit 736 determines that it is impossible to estimate future traffic volume using traffic estimation, and outputs information indicating the flow ID of the determined flow and the current traffic volume to the high-computational-complexity prediction unit 732f.

[0298] The large-computation-amount prediction unit 732f predicts future traffic volume using the large-computation-amount prediction and outputs the value of the future traffic volume predicted for each flow to the total traffic calculation unit 740.

[0299] The small calculation amount prediction unit 733 predicts future traffic volume using the small calculation amount prediction, and outputs the value of the future traffic volume predicted for each flow to the total traffic calculation unit 740.

[0300] The traffic estimation unit 734f estimates future traffic volume using traffic estimation. Here, the traffic estimation unit 734f performs traffic estimation using an algorithm that estimates future traffic volume based on fluctuations in the number of terminals within a traffic period. The traffic estimation unit 734f outputs the value of the future traffic volume estimated for each flow to the total traffic calculation unit 740.

[0301] The processing performed by the prediction execution unit 730f will be described below. Fig. 26 is a flowchart showing the operation of the prediction execution unit 730f of the communication system according to the sixth embodiment of the present invention.

[0302] The first prediction method determination unit 731f of the prediction execution unit 730f calculates the maximum value of the difference between the immediately preceding traffic volume and the current traffic volume for each flow classified by the flow classification unit 710. The prediction method determination unit 731f calculates the spread of traffic volumes (burstiness) based on the calculated maximum value (step S231).

[0303] Next, the first prediction method determination unit 731f determines whether the difference between the calculated traffic volumes is small. Specifically, the first prediction method determination unit 731f determines whether the maximum value of the difference between the traffic volume one period ago calculated in step S231 and the current traffic volume is smaller than a predetermined threshold (step S232).

[0304] If it is determined that the difference in traffic volume is small (Yes in step S232), the first prediction method determination unit 731f determines that future traffic volume can be predicted using low-computation-amount prediction. In this case, the low-computation-amount prediction unit 733 predicts future traffic volume using low-computation-amount prediction (step S233). The low-computation-amount prediction unit 733 outputs the value of the future traffic volume predicted for each flow to the total traffic calculation unit 740 (step S238).

[0305] On the other hand, if it is determined that the difference in traffic volume is not small (step S232: No), the first prediction method determination unit 731f calculates the traffic period of each flow. The traffic period calculation unit 735 outputs the calculated traffic period value of each flow to the prediction method determination unit 731d. The prediction method determination unit 731d acquires the traffic period value of each flow output from the traffic period calculation unit 735. The prediction method determination unit 731d calculates, for each flow classified by the flow classification unit 710, the number of terminals that has changed within the traffic period (step S234).

[0306] Next, the second prediction method determination unit 736 determines whether there has been a change in the number of terminals within the traffic cycle. Specifically, the second prediction method determination unit 736 determines whether the changed number of terminals calculated in step S234 above is 0 (step S235).

[0307] If it is determined that the number of terminals has not changed within the traffic cycle (No in step S235), the second prediction method determination unit 736 determines that future traffic volume can be estimated using traffic estimation. In this case, the traffic estimation unit 734f estimates future traffic volume using traffic estimation (step S236). The traffic estimation unit 734f outputs the value of the future traffic volume estimated for each flow to the total traffic calculation unit 740 (step S238).

[0308] On the other hand, if it is determined that the number of terminals has changed during the traffic cycle (step S235, Yes), the second prediction method determination unit 736 determines that it is impossible to estimate future traffic volume using traffic estimation. In this case, the computationally intensive prediction unit 732f predicts future traffic volume using computationally intensive prediction. The computationally intensive prediction unit 732f outputs the value of the future traffic volume predicted for each flow to the total traffic calculation unit 740 (step S238).

[0309] This completes the operation of the prediction execution unit 730f shown in the flowchart of FIG.

[0310] 25 , the total traffic calculation unit 740 acquires the values ​​of future traffic volume for each flow output from the large-computation-amount prediction unit 732 f, the small-computation-amount prediction unit 733, and the traffic estimation unit 734 f. The total traffic calculation unit 740 adds up the acquired values ​​of future traffic volume to obtain the value of total traffic volume. The total traffic calculation unit 740 outputs the calculated value of total traffic volume to the switching threshold determination unit 172.

[0311] The processing after the value of the total traffic volume is output from the total traffic calculation unit 740 of the prediction unit 171f to the switching threshold determination unit 172 (i.e., processing related to determining whether to perform route switching and processing related to controlling the implementation of route switching) is not described here because it is the same as the processing in the conventional communication system 1. In other words, the subsequent processing after the processing by the prediction unit 171f is the same as the processing related to the control of route switching by the switching threshold determination unit 172 and the switching determination unit 175 of the conventional communication system 1 described with reference to the above-mentioned Figures 9 to 12, and the processing related to route switching by the transfer device 15, and therefore not described here.

[0312] As described above, the switching instruction device 17f in the sixth embodiment of the present invention classifies each of the flows input from, for example, the lower-level device 3 during a predetermined period based on the characteristics of the traffic pattern (e.g., QoS) based on information defined in, for example, 5QI or S-NSSAI. For each classified flow, the switching instruction device 17f calculates the spread (burstiness) of traffic volume based on the time-series fluctuation (difference value) of the traffic volume. Based on the calculated spread (burstiness) of traffic volume, the switching instruction device 17f determines for each flow whether it is possible to predict future traffic volume using low-computational-load prediction, which is a prediction method with a relatively low computational load, such as linear prediction.

[0313] The switching instruction device 17f performs low-computational-amount prediction for a flow for which it is determined that future traffic volume can be predicted using low-computational-amount prediction. On the other hand, for a flow for which it is determined that future traffic volume cannot be predicted using low-computational-amount prediction, the switching instruction device 17f determines whether future traffic volume can be estimated using traffic estimation.

[0314] At this time, the prediction method determination unit 731d in this embodiment calculates the changed number of terminals for each traffic cycle obtained from the traffic cycle calculation unit 735. The prediction method determination unit 731c can perform traffic estimation using the traffic cycle, thereby reducing the amount of input data and further reducing the amount of calculation in the process of predicting future traffic volume.

[0315] The switching instruction device 17f performs traffic estimation for flows for which it is determined that future traffic volume can be estimated using traffic estimation. On the other hand, for flows for which it is determined that future traffic volume cannot be estimated using traffic estimation, the switching instruction device 17f performs high-computational-volume prediction, which is a prediction method with a relatively high computational volume, such as a prediction method using machine learning. The switching instruction device 17f adds up the future traffic volumes calculated for each flow by the high-computational-volume prediction, the low-computational-volume prediction, or the traffic estimation, and determines and controls the execution of route switching based on the total traffic volume.

[0316] As described above, the switching instruction device 17f according to the sixth embodiment of the present invention is configured to predict future traffic volume for each classified flow rather than predicting all flows collectively, thereby preventing degradation of communication quality. Furthermore, the switching instruction device 17f according to the sixth embodiment of the present invention is configured to perform low-computational-amount prediction, such as linear prediction, which can perform prediction with less computational effort, without performing high-computational-amount prediction using, for example, machine learning, for flows for which it is determined that future traffic volume can be predicted using low-computational-amount prediction. Furthermore, the switching instruction device 17f according to the sixth embodiment of the present invention is configured to perform traffic estimation, which can perform estimation with less computational effort, without performing high-computational-amount prediction using, for example, machine learning, for flows for which it is determined that future traffic volume can be estimated using traffic estimation. By providing such a configuration, the switching instruction device 17f can reduce the amount of computation required in predicting future traffic volume.

[0317] As a result, the switching instruction device 17f can reduce the amount of calculation required for predicting future traffic volume, thereby reducing communication delays that occur when prediction of future traffic volume and route switching are not completed before congestion occurs. In this way, the switching instruction device 17f in the sixth embodiment of the present invention can reduce communication delays while satisfying communication quality even when multiple flows are mixed.

[0318] As a result, the switching instruction device 17f can reduce the amount of calculation required for predicting future traffic volume, thereby reducing communication delays that occur when prediction of future traffic volume and route switching are not completed before congestion occurs. In this way, the switching instruction device 17f in the sixth embodiment of the present invention can reduce communication delays while satisfying communication quality even when multiple flows are mixed.

[0319] <Seventh Embodiment> A communication system according to a seventh embodiment of the present invention will now be described. Note that the overall configuration of the communication system according to this embodiment is similar to the configuration of the conventional communication system 1 shown in the overall configuration diagram of Fig. 1 described above, and therefore a description thereof will be omitted. Furthermore, the overall configuration of a mobile communication system in which the communication system according to this embodiment is applied to mobile communication is similar to the configuration of the conventional mobile communication system 10 shown in the overall configuration diagram of Fig. 2 described above, and therefore a description thereof will be omitted.

[0320] Furthermore, the block diagram showing the functional configuration of the switching instruction device of the communication system according to the seventh embodiment is the same as the block diagram showing the functional configuration of the switching instruction device 17 of the conventional communication system 1 shown in FIG. 3 described above. Hereinafter, the switching instruction device of the seventh embodiment will be referred to as a switching instruction device 17g. The switching instruction device 17g is, for example, an information processing device such as a general-purpose computer. The switching instruction device 17g of this embodiment and the conventional switching instruction device 17 described above differ in the configurations of the prediction unit, switching threshold determination unit, and switching determination unit. Below, the configurations of the prediction unit 171g, switching threshold determination unit 172g, and switching determination unit 175g of the communication system according to this embodiment will be described.

[0321] [Functional Configuration of Prediction Unit] Fig. 27 is a block diagram showing the functional configuration of a prediction unit 171g of a communication system according to the seventh embodiment of the present invention. As shown in Fig. 27, the prediction unit 171g includes a flow classification unit 710, a flow information storage unit 720, and a prediction execution unit 730g.

[0322] As in the first embodiment described above, the flow classification unit 710 classifies flows input to the network 5 during a predetermined period based on traffic characteristics. The flow classification unit 710 classifies flows by flow type based on information defined in, for example, 5QI or S-NSSAI. The flow classification unit 710 calculates the total number of terminals and traffic volume for each classified flow and stores them in a flow information table stored in the flow information storage unit 720.

[0323] The prediction execution unit 730g determines whether future traffic volume can be predicted using low-computation-amount prediction, which is a prediction method with a relatively low computational complexity, for each flow grouped (classified) by the flow classification unit 710. For flows for which it has determined that future traffic volume can be predicted using low-computation-amount prediction, the prediction execution unit 730g predicts future traffic volume using low-computation-amount prediction. Furthermore, for flows for which it has determined that future traffic volume cannot be predicted using low-computation-amount prediction, the prediction execution unit 730g further determines whether future traffic volume can be estimated using traffic estimation.

[0324] For flows for which it is determined that future traffic volume can be estimated using traffic estimation, the prediction execution unit 730g estimates future traffic volume using traffic estimation. For flows for which it is determined that future traffic volume cannot be estimated using traffic estimation, the prediction execution unit 730g predicts future traffic volume using high-computational-volume prediction, which is a prediction method with a relatively high computational load. The prediction execution unit 730g outputs the value of the predicted (or estimated) future traffic volume for each flow to the total traffic calculation unit 740.

[0325] As shown in FIG. 27, the prediction execution unit 730g is composed of a first prediction method determination unit 731g, a traffic period calculation unit 735g, a second prediction method determination unit 736, a large calculation amount prediction unit 732g, a small calculation amount prediction unit 733, and a traffic estimation unit 734g.

[0326] The first prediction method determination unit 731g determines whether it is possible to predict future traffic volume using low-computational-amount prediction, based on the traffic volume and traffic pattern, for each flow grouped (classified) by the flow classification unit 710. More specifically, the first prediction method determination unit 731g calculates the time-series fluctuation of traffic volume (traffic volume spread (burstiness)) for each flow grouped by the flow classification unit 710, and determines whether it is possible to predict future traffic volume using low-computational-amount prediction, based on the calculation result.

[0327] If the first prediction method determination unit 731g determines that future traffic volume can be predicted using low-computation-amount prediction, it outputs information indicating the flow ID of the determined flow and the current traffic volume to the low-computation-amount prediction unit 733. On the other hand, if the first prediction method determination unit 731g determines that future traffic volume cannot be predicted using low-computation-amount prediction, it outputs information indicating the flow ID of the determined flow and the current traffic volume to the traffic period calculation unit 735g.

[0328] The traffic period calculation unit 735g acquires information indicating the current traffic volume output from the first prediction method determination unit 731g. The traffic period calculation unit 735g calculates the traffic period of each flow. The traffic period calculation unit 735g outputs the calculated traffic period value of each flow and information indicating the current traffic volume to the second prediction method determination unit 736.

[0329] The second prediction method determination unit 736 acquires information indicating the traffic period value and current traffic volume of each flow output from the traffic period calculation unit 735g. The second prediction method determination unit 736 determines whether it is possible to estimate future traffic volume using traffic estimation for each flow grouped (classified) by the flow classification unit 710. More specifically, the second prediction method determination unit 736 calculates the changed number of terminals for each flow grouped by the flow classification unit 710, and determines whether it is possible to estimate future traffic volume using traffic estimation based on the calculation result.

[0330] At this time, the second prediction method determination unit 736 in this embodiment calculates the number of terminals that has changed for each traffic period obtained from the traffic period calculation unit 735 g. If the number of terminals that has changed within the traffic period is zero (i.e., if the number of terminals has not changed within the traffic period), the second prediction method determination unit 736 determines that it is possible to estimate future traffic volume using traffic estimation, and outputs information indicating the flow ID of the determined flow and the current traffic volume to the traffic estimation unit 734 g.

[0331] On the other hand, if the number of terminals that has changed within the traffic cycle is one or more (i.e., if the number of terminals has changed within the traffic cycle), the second prediction method determination unit 736 determines that it is impossible to estimate future traffic volume using traffic estimation, and outputs information indicating the flow ID of the determined flow and the current traffic volume to the high-computational-volume prediction unit 732g.

[0332] The large-computation-amount prediction unit 732g predicts future traffic volume using the large-computation-amount prediction and outputs the value of the future traffic volume predicted for each flow to a total traffic calculation unit 740g of the switching threshold determination unit 172g, which will be described later.

[0333] The small-computation-amount prediction unit 733 predicts future traffic volume using the small-computation-amount prediction and outputs the value of the future traffic volume predicted for each flow to a total traffic calculation unit 740g of the switching threshold determination unit 172g, which will be described later.

[0334] The traffic estimation unit 734g estimates future traffic volume using traffic estimation. Here, the traffic estimation unit 734f performs traffic estimation using an algorithm that estimates future traffic volume based on fluctuations in traffic volume associated with traffic cycles. The traffic estimation unit 734g outputs the value of the future traffic volume estimated for each flow to a total traffic calculation unit 740g of the switching threshold determination unit 172g, which will be described later.

[0335] The prediction units in the first to sixth embodiments described above include a total traffic calculation unit 740, and the large calculation amount prediction unit, the small calculation amount prediction unit, and the traffic prediction unit are configured to output predicted (or estimated) future traffic volume values ​​to the total traffic calculation unit 740. The total traffic calculation unit 740 is configured to acquire and add up the acquired future traffic volume values, and output the total traffic volume value to the switching threshold determination unit.

[0336] In contrast to this, in the seventh embodiment, the prediction unit 171g does not include a total traffic calculation unit 740, and instead, a switching threshold determination unit 172g (described later) includes a total traffic calculation unit 740g.

[0337] The switching instruction devices of the first to sixth embodiments described above and the switching instruction device of the seventh embodiment have different path switching methods. Specifically, the switching instruction devices of the first to sixth embodiments described above are configured to perform path switching for each lower-level device 3 based on the predicted traffic volume, but the switching instruction device of the seventh embodiment is configured to perform path switching for each flow based on the predicted traffic volume.

[0338] [Configuration of Switching Threshold Determination Unit and Switching Determination Unit] The following describes the configurations of the switching threshold determination unit 172g and the switching determination unit 175g of the switching instruction device 17g in the seventh embodiment. Figure 28 is a block diagram showing the functional configurations of the switching threshold determination unit 172g and the switching determination unit 175g of the switching instruction device 17g in the seventh embodiment of the present invention.

[0339] As shown in FIG. 28, the switching threshold value determination unit 172 g includes a total traffic calculation unit 740 g , a prediction accuracy calculation unit 173 , and a threshold value calculation unit 174 .

[0340] The total traffic calculation unit 740g acquires values ​​of future traffic volume predicted by the high-computational-amount prediction unit 732g, the low-computational-amount prediction unit 733, and the traffic estimation unit 734g of the prediction execution unit 730. The total traffic calculation unit 740g calculates the total traffic volume of actual traffic per lower-level device 3 (hereinafter referred to as the "total actual traffic volume") and the total traffic volume per lower-level device 3 (hereinafter also referred to as the "total predicted traffic volume"). The total traffic calculation unit 740g outputs the calculated values ​​of the total actual traffic volume and the total predicted traffic volume to the prediction accuracy calculation unit 173.

[0341] The prediction accuracy calculation unit 173 acquires the value of the total actual traffic volume and the value of the total predicted traffic volume output from the total traffic calculation unit 740g. The prediction accuracy calculation unit 173 calculates the prediction accuracy for each upper link based on the future traffic volume predicted in the past by the prediction unit 171 and the total actual traffic volume. The prediction accuracy calculation unit 173 outputs information indicating the calculated prediction accuracy to the threshold calculation unit 174.

[0342] The threshold calculation unit 174 acquires the information indicating the prediction accuracy output from the prediction accuracy calculation unit 173. The threshold calculation unit 174 calculates a switching threshold for each upper link based on the link rate and the prediction accuracy. The threshold calculation unit 174 outputs the switching threshold for each upper link to the switching determination unit 175g.

[0343] As shown in FIG. 28, the switching determination unit 175 g includes a link usage rate calculation unit 176 and a switching flow calculation unit 177 .

[0344] The link utilization calculation unit 176 acquires the switching threshold for each upper link output from the threshold calculation unit 174. The link utilization calculation unit 176 also acquires the value of the total predicted traffic volume from the total traffic calculation unit 740g. The link utilization calculation unit 176 calculates the predicted link utilization for each upper link based on the value of the total predicted traffic volume and the switching threshold. The link utilization calculation unit 176 outputs the calculated link utilization value to the switching flow calculation unit 177.

[0345] The switching flow calculation unit 177 acquires the link utilization value output from the link utilization calculation unit 176. When there is an upper link for which it is determined that congestion will occur due to a high predicted link utilization rate, the switching flow calculation unit 177 determines to perform route switching, and identifies route switching that will result in a traffic volume that will not cause congestion, starting with the flow with the lowest priority. When it determines to perform route switching, the switching flow calculation unit 177 instructs each device (e.g., the lower device 3, the upper device 4, and the transfer device 15) to switch the identified route.

[0346] Hereinafter, a more detailed explanation will be given, using a specific example, of an example of the operation of the switching flow calculation unit 177. Fig. 29 is a flowchart showing the operation of the switching flow calculation unit 177 of the switching determination unit 175g in the seventh embodiment of the present invention.

[0347] The switching flow calculation unit 177 obtains the future traffic volume and sorts the predicted traffic volume for each flow. The switching flow calculation unit 177 sorts the flows in ascending order of priority (step S241).

[0348] In addition, the switching flow calculation unit 177 calculates the congestion amount (or the free space if the value is negative) for each lower-level device 3 based on the link usage rate of each lower-level device 3 and the switching threshold, as shown in the following equation (21) (step S242).

[0349] X_conv=(link utilization rate−1)×switching threshold (21)

[0350] Hereinafter, X_conv where the congestion amount is a positive value will be represented as "X_conv1", and X_conv where the congestion amount is a negative value will be represented as "X_conv2".

[0351] The switching flow calculation unit 177 identifies the lower-level device 3 with the largest congestion amount value and the lower-level device 3 with the smallest congestion amount value (i.e., the largest available capacity value). Based on the largest congestion amount value and the largest available capacity value, the switching flow calculation unit 177 calculates the maximum traffic volume at which path switching is effective (step S243).

[0352] For example, if the congestion level of the first lower-level device 3 is the highest and the congestion level of the second lower-level device 3 is the lowest (i.e., the available capacity is the highest), the maximum traffic volume at which path switching is effective is expressed by the following equation (22).

[0353] X=X_conv1−X_conv2…(22)

[0354] The switching flow calculation unit 177 determines whether the route switching is effective by switching the routes in order from the lowest priority flow (Step S244).

[0355] If it is determined that the route switching is effective (X>Xi) (Yes in step S244), the switching flow calculation unit 177 instructs the link usage rate calculation unit 176 to recalculate the link usage rate (Link).

[0356] Link = Link-xi / switching threshold (23)

[0357] The switching flow calculation unit 177 changes the lower-level device 3 to which the switched predicted traffic belongs (Step S245).

[0358] On the other hand, if it is determined that the path switching is ineffective (X≦Xi) (No in step S244), the switching flow calculation unit 177 compares the flow with the next highest priority (step S247). Note that in this embodiment, the comparison is performed in order from the lowest priority flow, but the present invention is not limited to this configuration. For example, the comparison may be performed in order from the highest traffic volume flow.

[0359] When performing path switching for each flow, the switching flow calculation unit 177 notifies the transfer device 15 (switching device) of the DSCP of the flow for which path switching is to be performed, since there is a correspondence between the DSCP (Differentiated Services Code Point) and 5QI.

[0360] After determining whether to perform route switching, the switching flow calculation unit 177 again determines whether congestion has occurred based on the new link utilization rates. If all link utilization rates are not 1 (i.e., if the link utilization rates are greater than 1), the switching flow calculation unit 177 ends the process.

[0361] This completes the operation of the switching flow calculation unit 177 shown in the flowchart of FIG.

[0362] As described above, the switching instruction device 17g in the seventh embodiment of the present invention classifies each of the flows input from, for example, the lower-level device 3 during a predetermined period based on the characteristics of the traffic pattern (e.g., QoS) based on information defined in, for example, 5QI or S-NSSAI. For each classified flow, the switching instruction device 17g calculates the spread (burstiness) of traffic volume based on the time-series fluctuation (difference value) of the traffic volume. Based on the calculated spread (burstiness) of traffic volume, the switching instruction device 17g determines for each flow whether it is possible to predict future traffic volume using low-computational-load prediction, which is a prediction method with a relatively low computational load, such as linear prediction.

[0363] The switching instruction device 17g performs low-computational-amount prediction for a flow for which it is determined that future traffic volume can be predicted using low-computational-amount prediction. On the other hand, for a flow for which it is determined that future traffic volume cannot be predicted using low-computational-amount prediction, the switching instruction device 17g determines whether future traffic volume can be estimated using traffic estimation.

[0364] At this time, the second prediction method determination unit 736 in this embodiment calculates the changed number of terminals for each traffic cycle obtained from the traffic cycle calculation unit 735. The second prediction method determination unit 736 can perform traffic estimation using the traffic cycle, thereby reducing the amount of input data and further reducing the amount of calculation in the process of predicting future traffic volume.

[0365] The switching instruction device 17g performs traffic estimation for flows for which it is determined that future traffic volume can be estimated using traffic estimation. On the other hand, for flows for which it is determined that future traffic volume cannot be estimated using traffic estimation, the switching instruction device 17g performs high-computational-volume prediction, which is a prediction method with a relatively high computational volume, such as a prediction method using machine learning. The switching instruction device 17g sums up the future traffic volumes calculated for each flow by the high-computational-volume prediction, the low-computational-volume prediction, or the traffic estimation, and determines and controls the execution of route switching based on the total traffic volume.

[0366] As described above, the switching instruction device 17g according to the seventh embodiment of the present invention is configured to predict future traffic volume for each classified flow rather than predicting all flows collectively, thereby preventing degradation of communication quality. Furthermore, the switching instruction device 17g according to the seventh embodiment of the present invention is configured to perform low-computational-amount prediction, such as linear prediction, which can perform prediction with less computational effort, without performing high-computational-amount prediction using, for example, machine learning, for flows for which it is determined that future traffic volume can be predicted using low-computational-amount prediction. Furthermore, the switching instruction device 17g according to the seventh embodiment of the present invention is configured to perform traffic estimation, which can perform estimation with less computational effort, without performing high-computational-amount prediction using, for example, machine learning, for flows for which it is determined that future traffic volume can be estimated using traffic estimation. By providing such a configuration, the switching instruction device 17g can reduce the amount of computation required in predicting future traffic volume.

[0367] As a result, the switching instruction device 17g can reduce the amount of calculation required for predicting future traffic volume, thereby reducing communication delays that occur when prediction of future traffic volume and route switching are not completed before congestion occurs. In this way, the switching instruction device 17g in the seventh embodiment of the present invention can reduce communication delays while satisfying communication quality even when multiple flows are mixed.

[0368] Hereinafter, a description will be given of a determination method used by the prediction method determination unit and the first prediction method determination unit in each of the above-described embodiments to determine whether it is possible to predict future traffic volume using low-computational-amount prediction. Also, a description will be given of a determination method used by the prediction method determination unit and the second prediction method determination unit in each of the above-described embodiments to determine whether it is possible to estimate future traffic volume using traffic estimation.

[0369] Fig. 30 is a diagram for explaining the determination of a prediction method by a prediction method determination unit in each embodiment of the present invention. Fig. 30 shows four graphs, flow 1 to flow 4. In each graph, the vertical axis represents traffic volume and the horizontal axis represents time. That is, Fig. 30 is a diagram showing an example of the transition of traffic volume over time for four types of flows. Note that, here, it is assumed that traffic volume is proportional to the number of terminals that are traffic sources.

[0370] As shown in Figure 30, flows 1 and 4 are characterized by irregular changes in traffic volume over time. Flows 2 and 3 are characterized by alternating traffic occurrences at regular intervals (i.e., burstiness). Flow 2 is characterized by an inconstant traffic volume when traffic occurs (i.e., the number of terminals changes), while flow 3 is characterized by a constant traffic volume when traffic occurs (i.e., the number of terminals does not change).

[0371] For example, in the case where there are four types of flows, the prediction method determination unit in each embodiment of the present invention determines the prediction method based on the following rules.

[0372] For example, when future traffic volume for each flow is predicted using high-computational-complexity prediction, as in conventional communication systems, the computational load associated with the prediction increases in proportion to the number of flows, lengthening the computation time required for the prediction and making it impossible to complete route switching before congestion actually occurs. Therefore, the prediction method determination unit in each embodiment of the present invention reduces the computational complexity in the process of predicting future traffic volume by using the low-computational-complexity prediction or the traffic estimation described above instead of the high-computational-complexity prediction, depending on traffic characteristics such as the traffic volume and traffic pattern of each flow.

[0373] For example, the prediction method determination unit determines to perform low-computational-load prediction instead of high-computational-load prediction because low-computational-load prediction such as linear prediction can be used for flows that have a characteristic that the traffic volume changes continuously over time, such as flows 1 and 4. Note that for flows with changing traffic volume such as flows 1 and 4, even if high-computational-load prediction using machine learning or the like is performed, accuracy is not guaranteed.

[0374] Furthermore, for example, the prediction method determination unit determines not to perform low-computation-amount prediction, for example, linear prediction, for flows such as flow 2 and flow 3, whose traffic volume changes discretely over time (i.e., flows with burstiness), because low-computation-amount prediction cannot be used. In this case, the prediction method determination unit checks whether or not there is a change in the number of terminals per traffic cycle (i.e., traffic volume per traffic cycle) that are the source of traffic.

[0375] For example, for a flow such as flow 3, which has the characteristic that the number of terminals does not change for each traffic period, the prediction method determination unit determines to perform traffic estimation instead of high-computational-volume prediction, because, for example, the traffic volume can be estimated (i.e., past traffic volume can be considered as future traffic volume).

[0376] On the other hand, for example, for flows such as flow 2, which have the characteristic that the number of terminals changes with each traffic cycle, the prediction method determination unit determines that a computationally intensive prediction should be performed using, for example, machine learning, because past traffic volume cannot be considered as future traffic volume.

[0377] In this way, the prediction method determination unit in each of the above-described embodiments can reduce the number of flows for which calculation-intensive prediction is performed, and reduce the amount of calculation in the process of predicting future traffic volume.

[0378] The first to third embodiments are described assuming the use of traffic captured from the lower-level device 3. In each of the above-described embodiments, the network 5 may be any network such as a mobile fronthaul (MFH), a mobile midhaul (MMH), or a mobile backhaul (MBH).

[0379] In each of the above-described embodiments, the switching instruction device may be configured to acquire traffic information from a controller (not shown) of the lower-level device 3 instead of acquiring traffic information from the lower-level device 3 .

[0380] In each of the above-described embodiments, the switching instruction device may be configured to use scheduling information in the wireless section (e.g., BSR (Buffer Status Report), TBS (Transport Block Size), etc.) instead of using a traffic volume value when determining whether to perform path switching. In this case, the network 5 is, for example, an MMH or MBH. The source from which the scheduling information is obtained is, for example, a DU, a DU controller, a CU, or a CU controller.

[0381] In the above-described embodiments, the case of upstream communication has been described as an example, but the present invention is not limited to this, and a similar configuration can be applied to downstream communication as well.

[0382] In the above-described embodiments, a configuration assuming a 5G mobile network is described, but the present invention is not limited to this. A similar configuration can be applied to other wireless networks such as Wi-Fi (registered trademark). In this case, the source from which traffic information is obtained is, for example, a Wi-Fi (registered trademark) access point. Furthermore, the network 5 is, for example, a network between a Wi-Fi (registered trademark) controller and a Wi-Fi (registered trademark) access point.

[0383] As described above, the switching instruction device in each of the above-described embodiments of the present invention determines, for each classified flow, whether low-computation-amount prediction or traffic estimation, which requires relatively less computational effort than high-computation-amount prediction, such as a prediction method using machine learning, is possible. The switching instruction device is then configured to perform low-computation-amount prediction or traffic estimation without performing high-computation-amount prediction for flows for which it is determined that low-computation-amount prediction or traffic estimation is possible. This allows the switching instruction device to reduce the amount of computation required for predicting future traffic volume, thereby reducing communication delays caused by not completing future traffic volume prediction and route switching before congestion occurs. This allows the switching instruction device in each of the above-described embodiments of the present invention to reduce communication delays while maintaining communication quality, even when multiple flows are mixed.

[0384] According to the above-described embodiment, the prediction device includes a flow classification unit, a first prediction unit, a second prediction unit, a prediction method determination unit, and a total traffic calculation unit. For example, the prediction device is 171a to 171g in the embodiment, the first prediction unit is the large-computation-amount prediction unit 732, 732e to 732g in the embodiment, the second prediction unit is the small-computation-amount prediction unit 733 or the traffic estimation unit 734, 734d, 734f, 734g in the embodiment, the prediction method determination unit is the prediction method determination unit 731, 731b to 731e in the embodiment, and the total traffic calculation unit is the total traffic calculation unit 740, 740g in the embodiment.

[0385] The flow classification unit classifies communication traffic input during a predetermined period into a plurality of flows based on characteristics of the communication traffic. The first prediction unit predicts future traffic volumes of the flows using a first prediction method based on current traffic volumes of the flows. For example, the first prediction method is high-computational complexity prediction in the embodiment. The second prediction unit predicts future traffic volumes of the flows using a second prediction method that requires less computational complexity than the first prediction method based on current traffic volumes of the flows. For example, the second prediction method is low-computational complexity prediction or traffic estimation in the embodiment. The prediction method determination unit determines the prediction method to be used for each flow based on characteristics of the communication traffic. The total traffic calculation unit calculates the sum of the traffic volumes for each flow predicted by the first prediction unit and the traffic volumes for each flow predicted by the second prediction unit.

[0386] In the above prediction device, the prediction method determination unit may determine to use the second prediction method when the difference between the traffic volume of the flow in the most recent period and the traffic volume of the flow in the immediately preceding period is smaller than a predetermined threshold. In this case, for example, the prediction device is the prediction unit 171 in the first embodiment, the prediction method determination unit is the prediction method determination unit 731 in the first embodiment, the difference is the value of Traffic_burst in the first embodiment, and the predetermined threshold is the value of T_th in the first embodiment.

[0387] In the above prediction device, the prediction method determination unit may determine to use the second prediction method when the average value of the traffic volume of the flow in the most recent period is smaller than a predetermined threshold. In this case, for example, the prediction device is the prediction unit 171b in the second embodiment, the prediction method determination unit is the prediction method determination unit 731b in the second embodiment, the average value is the value of Traffic_ave in the second embodiment, and the predetermined threshold is the value of T_th in the second embodiment.

[0388] In the above prediction device, the prediction method determination unit may determine not to use the first prediction method if there is no change in the number of terminals that are source of communication traffic. In this case, the prediction device is the prediction unit 171c in the third embodiment and the prediction unit 171d in the fourth embodiment, the prediction method determination unit is the prediction method determination unit 731c in the third embodiment and the prediction method determination unit 731d in the fourth embodiment, and the number of terminals is the value of Num_val in the third and fourth embodiments.

[0389] In the prediction device described above, the prediction method determination unit may determine not to use the first prediction method if there has been no change in the number of terminals in the most recent period. In this case, the prediction device is the prediction unit 171d in the fourth embodiment, the prediction method determination unit is the prediction method determination unit 731d in the fourth embodiment, and the number of terminals is the value of Num_val in the fourth embodiment.

[0390] The prediction device may further include a traffic period calculation unit. In this case, for example, the prediction device is the prediction unit 171d in the fourth embodiment, the prediction unit 171e in the fifth embodiment, the prediction unit 171f in the sixth embodiment, and the prediction unit 171g in the seventh embodiment, and the traffic period calculation unit is the traffic period calculation unit 735 in the fourth and fifth embodiments, the traffic period calculation unit 735f in the sixth embodiment, and the traffic period calculation unit 735g in the seventh embodiment. In this case, the traffic period calculation unit calculates the period of communication traffic based on the traffic volume. Also, in this case, the first prediction unit predicts future traffic volume for each period calculated by the traffic period calculation unit.

[0391] Furthermore, according to the above-described embodiment, the communication control device includes a flow classification unit, a first prediction unit, a second prediction unit, a prediction method determination unit, a total traffic calculation unit, a switching determination unit, and a switching instruction unit. For example, the prediction devices are 171a to 171g in the embodiment, the first prediction unit is the large-computation-amount prediction unit 732, 732e to 732g in the embodiment, the second prediction unit is the small-computation-amount prediction unit 733 or the traffic estimation unit 734, 734d, 734f, 734g in the embodiment, the prediction method determination unit is the prediction method determination unit 731, 731b to 731e in the embodiment, the total traffic calculation unit is the total traffic calculation unit 740, 740g in the embodiment, the switching determination unit is the switching determination unit 175 and 175g in the embodiment, and the switching instruction unit is the switching instruction device 17a to 17g in the embodiment.

[0392] The flow classification unit classifies communication traffic input during a predetermined period into multiple flows based on characteristics of the communication traffic. The first prediction unit predicts future traffic volumes of the flows using a first prediction method based on current traffic volumes of the flows. For example, the first prediction method is the high-computational-amount prediction in the embodiment. The second prediction unit predicts future traffic volumes of the flows using a second prediction method that requires less computational effort than the first prediction method based on current traffic volumes of the flows. For example, the second prediction method is the low-computational-amount prediction or traffic estimation in the embodiment. The prediction method determination unit determines a prediction method to be used for each flow based on characteristics of the communication traffic. The total traffic calculation unit calculates the sum of the traffic volumes for each flow predicted by the first prediction unit and the traffic volumes for each flow predicted by the second prediction unit. The switching determination unit predicts whether congestion will occur in the future based on the sum and determines whether a communication path switch is necessary based on the prediction result. The switching instruction unit issues a switching instruction to a switching device that executes communication path switch when the switching determination unit determines that a communication path switch is necessary. For example, the switching device is the transfer device 15 in the embodiment.

[0393] Part or all of the configuration of the switching instruction device in the above-described embodiment may be implemented by a computer. In this case, a program for implementing this function may be recorded on a computer-readable recording medium, and the program may be loaded into a computer system and executed. Note that the term "computer system" as used herein includes hardware such as an OS and peripheral devices. Furthermore, "computer-readable recording medium" refers to portable media such as flexible disks, optical magnetic disks, ROMs, and CD-ROMs, as well as storage devices such as hard disks built into computer systems. Furthermore, "computer-readable recording medium" may also include media that dynamically store programs for a short period of time, such as communication lines used when transmitting programs via networks such as the Internet or telephone lines, or media that store programs for a fixed period of time, such as volatile memory within the computer system that serves as the server or client. The program may be designed to implement part of the above-described functions, or may be capable of implementing the above-described functions in combination with programs already stored in the computer system, or may be implemented using a programmable logic device such as an FPGA (Field Programmable Gate Array).

[0394] Although an embodiment of the present invention has been described in detail above with reference to the drawings, the specific configuration is not limited to this embodiment, and includes designs within the scope of the present invention that do not deviate from the gist of the present invention.

[0395] REFERENCE SIGNS LIST 1...communication system, 3...lower-level device, first lower-level device, second lower-level device, 3-1 to 3-4...lower-level devices, 4, 4-1 to 4-2...higher-level devices, 5...network, 7...switching instruction device, 10...mobile communication system, 11...terminal station, 12, 12-m...antenna station, 13, 13-1, 13-m, 3-M...remote station, 14, 14-1, 14-n, 14-N...aggregate station, 15...transfer device, 16...resource allocation device, 17, 17a to 17g...switching instruction device, 20...higher-level network, 71...prediction unit, 72...switching threshold determination unit, 73...prediction accuracy calculation unit, 74...threshold calculation unit, 75...switching judgment unit, 171, 171a to 171g...prediction unit, 172, 172g...switching threshold determination unit , 173... prediction accuracy calculation unit, 174... threshold calculation unit, 175, 175g... switching determination unit, 176... link utilization calculation unit, 177... switching flow calculation unit, 710... flow classification unit, 720... flow information storage unit, 729... flow information storage unit, 730, 730b to 730g... prediction execution unit, 731, 731b to 731e... prediction method determination unit, 731f, 731g... first prediction method determination unit, 732, 732e to 732g... high calculation amount prediction unit, 733... low calculation amount prediction unit, 734, 734d, 734f, 734g... traffic estimation unit, 735, 735f, 735g... traffic period calculation unit, 736... second prediction method determination unit, 740, 740g... total traffic calculation unit

Claims

1. A prediction device comprising: a flow classification unit that classifies communication traffic input during a specified period into a plurality of flows based on characteristics of the communication traffic; a first prediction unit that predicts a future traffic volume of the flow using a first prediction method based on a current traffic volume of the flow; a second prediction unit that predicts a future traffic volume of the flow using a second prediction method that requires less calculations than the first prediction method, based on the current traffic volume of the flow; a prediction method determination unit that determines a prediction method to be used for each of the flows based on the characteristics of the communication traffic; and a total traffic calculation unit that calculates the sum of the traffic volume for each of the flows predicted by the first prediction unit and the traffic volume for each of the flows predicted by the second prediction unit.

2. The prediction device according to claim 1, wherein the prediction method determination unit determines to use the second prediction method when a difference value between the traffic volume of the flow in the most recent period and the traffic volume of the flow in the immediately previous period is smaller than a predetermined threshold value.

3. The prediction device according to claim 1, wherein the prediction method determination unit determines to use the second prediction method when an average value of the traffic volume of the flow in the most recent period is smaller than a predetermined threshold value.

4. The prediction device according to claim 1, wherein the prediction method determination unit determines not to use the first prediction method when there is no change in the number of terminals that are the senders of the communication traffic.

5. The prediction device according to claim 4, wherein the prediction method determination unit determines not to use the first prediction method if there has been no change in the number of terminals in the most recent period.

6. A prediction device as claimed in any one of claims 1 to 5, further comprising a traffic period calculation unit that calculates a period of the communication traffic based on the traffic volume, wherein the first prediction unit predicts the future traffic volume for each of the periods calculated by the traffic period calculation unit.

7. A communication control device comprising: a flow classification unit that classifies communication traffic input during a predetermined period into a plurality of flows based on characteristics of the communication traffic; a first prediction unit that predicts a future traffic volume of the flow using a first prediction method based on a current traffic volume of the flow; a second prediction unit that predicts a future traffic volume of the flow using a second prediction method having a smaller amount of calculation than the first prediction method based on the current traffic volume of the flow; a prediction method determination unit that determines a prediction method to be used for each of the flows based on the characteristics of the communication traffic; a total traffic calculation unit that calculates a sum of the traffic volume for each of the flows predicted by the first prediction unit and the traffic volume for each of the flows predicted by the second prediction unit; a switching determination unit that predicts whether congestion will occur in the future based on the sum and determines whether or not switching of a communication path is necessary based on a prediction result; and a switching instruction unit that issues a switching instruction to a switching device that executes switching of the communication path when it is determined by the switching determination unit that switching of the communication path is necessary.

8. A prediction method executed by a computer, comprising: a flow classification step of classifying communication traffic input during a predetermined period into a plurality of flows based on characteristics of the communication traffic; a first prediction step of predicting a future traffic volume of the flow using a first prediction method based on a current traffic volume of the flow; a second prediction step of predicting a future traffic volume of the flow using a second prediction method requiring less calculations than the first prediction method based on the current traffic volume of the flow; a prediction method determination step of determining a prediction method to be used for each of the flows based on the characteristics of the communication traffic; and a total traffic calculation step of calculating the sum of the traffic volume for each of the flows predicted in the first prediction step and the traffic volume for each of the flows predicted in the second prediction step.

Citation Information

Patent Citations

  • Communication traffic predicting device, communication traffic predicting method, and program

    JP2014087031A

  • Routing control system, routing control method, and program

    JP2019169783A