Traffic Scheduling Method for Multipath Concurrent Transmission System Based on Fountain Code

Abstract

A traffic scheduling method for a multi-path concurrent transmission system based on fountain codes. The multi-path concurrent transmission system includes network devices such as a software defined network (SDN) controller, terminal nodes (transmitter, receiver), and network nodes (switches, routers). The terminal nodes periodically collect service flow status information and channel status information, and upload them to the SDN controller as the basis for the SDN controller to make scheduling decisions. The SDN controller dynamically optimizes the traffic scheduling strategy based on the current service flow information and channel status information, the principle of service fairness and the principle of minimum overall transmission delay, and publishes it to the transmitter. The transmitter proportionally schedules data packets to multiple network interfaces for transmission based on the optimal scheduling ratio issued by the SDN controller. The present invention can reduce the overall transmission delay of service flows while ensuring the fairness of multiple services, and achieve efficient use of network resources.

Description

Technical Field

[0001] The present invention relates to the field of heterogeneous multi-path transmission networks, and in particular to a flow scheduling method for a multi-path concurrent transmission system based on fountain codes. Background Art

[0002] With the enhancement of wireless transmission system functions and the diversification of services, communication traffic is growing and more attention is paid to data timeliness. Traditional single-link transmission technology is increasingly difficult to meet the different requirements of different services for bandwidth, delay and reliable information transmission in the future. Therefore, multipath concurrent transmission technology has become a research hotspot in the field of wireless communication. It enriches network diversity in heterogeneous wireless networks and accelerates the efficiency of content distribution. A more representative multipath concurrent transmission technology is the dual connection technology proposed in the 12th version of the 3GPP specification. Its user terminal can simultaneously connect to the macro base station and the small base station. The two base stations have different frequency bands, so the two channels are independent of each other and will not interfere with each other. In this system, a complete data stream will be split into two sub-data streams, which are transmitted simultaneously on two network interfaces and sent to the terminal. Therefore, the bandwidth of the two channels can be fully utilized for data transmission, thereby achieving lower transmission delay. Due to the introduction of redundant links, users can maintain connection with the network for a long time, avoiding the link interruption problem faced by traditional single connections when the signal is weak or the channel is switched.

[0003] However, the interference and instability of wireless connections in complex electromagnetic environments are difficult to predict in dynamic environments, resulting in great differences in transmission capacity, coverage, QoS support, and other aspects of multiple links of different communication technologies. At the same time, the imbalance of links will also cause the disorder of data packets at the receiving end, thereby bringing additional network delays. Therefore, multipath concurrent transmission requires the design of sophisticated flow control and diversion strategies at the sending end to solve problems such as data packet reordering due to different paths and frequent packet loss due to unreliable wireless channels. In the study of traditional multipath transmission or multi-access networks, some optimization resource scheduling schemes at different network protocol layers have been continuously proposed. The paper "Downlink tra c scheduling for LTE-A small cell networks with dual connectivity enhancement, IEEE Commun. Lett., 2016" proposes a downlink traffic scheduling scheme that takes into account factors such as the channel quality of users accessing macro base stations and small base stations, data transmission interval time, and base station cache data. During the transmission process, the diversion ratio between macro base stations and small base stations is dynamically adjusted to achieve better throughput. The paper "Online packet dispatching for delayoptimal concurrent transmissions in heterogeneous multi-RAT networks, IEEE Trans. Wireless Commun, 2016" considers the online packet scheduling problem of heterogeneous network multi-access systems and proposes a packet scheduling strategy based on link state, which can allocate packets to different access base stations according to traffic load and link state to reduce the packet reordering delay at the user end. However, the above schemes all require very detailed link state information, so the signaling overhead is large and the control complexity is extremely high. In addition to solving the diversion control in multi-path transmission and the disorder of receiving end packets from the perspective of traditional resource allocation, many studies in recent years have used network coding to solve related problems. For example, the paper "Network coding based multipath TCP, IEEE INFOCOM workshop 2012" studied the transmission balance problem caused by different network conditions of each path in the traditional multi-path transmission control protocol, and proposed a new multi-path transmission control model, which uses network coding schemes to encode packets on several paths before sending them; in addition to receiving regular data, the receiving end also receives network coded packets as redundant data to compensate for lost packets to reduce delay. However, this scheme is mainly based on the transport layer and needs to solve the problems of long round-trip time and multi-hop routing. Summary of the invention

[0004] In order to overcome the deficiencies of the above-mentioned prior art, the present invention proposes a multi-path concurrent transmission traffic scheduling method based on fountain codes. At the transmitting end, a new fountain code RaptorQ code is used to generate enough coded data for the original batch of data packets. At the receiving end, only enough coded data packets need to be received to decode the coded data packets and restore the original data, thereby effectively eliminating the problem of data packet disorder, reducing the complexity of multi-channel control, and reducing the delay caused by traditional packet loss retransmission and reordering. The present invention takes the encoding and decoding process of a batch of data packets in a multi-connection concurrent transmission system based on RaptorQ code as the research object, establishes a business traffic optimization model in combination with the relationship between various variables, and then jointly optimizes the traffic scheduling ratio on each channel under multi-business scenarios, which greatly simplifies the optimization model of traffic scheduling. After optimization, the concurrent transmission system can achieve the minimum overall transmission delay while ensuring the service fairness of different business data streams. The method comprises the following steps:

[0005] Step 1. The sender obtains the number of original data packets of the service flow received locally and encodes them, and reports the encoding length of each service flow to the SDN controller.

[0006] Step 2. The receiving end periodically obtains the status information of a total of m channels, including the propagation delay d of channel i (i = 1, ..., m) i , transmission success rate r i (the probability of successfully transmitting a data packet within a unit time slot), etc., and reported to the SDN controller.

[0007] Step 3. The SDN controller calculates the optimal service scheduling ratio for multi-channel concurrent transmission based on the service flow information fed back by the sender and the channel state information fed back by the receiver, in accordance with the service fairness principle and the principle of maximizing channel utilization.

[0008] Step 4: The SDN controller sends the optimal service scheduling ratio to the sending end.

[0009] Step 5. The sender dispatches the service according to the service scheduling ratio ρ issued by the SDN controller j , j = 1, ..., n, schedule the encoded data packet to the corresponding channel (network interface) and send it.

[0010] Step 6. The receiving end receives the coded data packets of each service stream transmitted via m channels. When receiving N packets of the jth service, j After a data packet is decoded, the K corresponding to the service is obtained. j The original data packet is sent to the sender, and the ACK corresponding to service j is sent to the sender j .

[0011] Step 7. The sender detects the received ACK status of each service flow. j After (j=1,...,n), the transmission ends.

[0012] In step 1, the sending end obtains local service flow status information and determines the encoding length of each service flow, and uploads the corresponding information to the SDN controller. The specific steps are as follows:

[0013] (1-1) The sender periodically obtains the relevant status information of the local n service flows, including the number of original data packets of each service in the current cache queue K j ,j=1,...,n (i.e., the length of the cache queues of multiple services), the waiting time of each service q j ;

[0014] (1-2) The sender determines the encoding length N of each service flow according to the original data packet length of each service flow and the RaptorQ standard. j (N j Indicates that the receiving end can successfully decode and restore the original batch of K j-th service flow j The number of coded data packets required for each data packet) and encode the original data packet. j Report to the SDN controller.

[0015] In step 2, the receiving end periodically obtains channel state information such as channel propagation delay and transmission success rate, and reports it to the SDN controller. The specific steps are as follows:

[0016] (2-1) Before forwarding the coded data packet to the i-th (i=1,...,m) channel, the sender writes fields for channel measurement in the header of each coded data packet, including: a). link number link_ID to which the data packet is forwarded; b). send batch number Send_batch_ID; c). send time T_send; d). send number send_count, which is reset to 1 when a new batch of coded data packets starts to be sent.

[0017] (2-2) The receiving end counts the received coded data packets, parses the corresponding fields in the packet header such as the data packet batch number, sending time and sending number, and calculates the average delay of the i-th link according to the following formula:

[0018]

[0019] In the above formula, sum_receive is the total number of packets received in the same batch, T l receive is the time when the lth data packet arrives at the receiving end, Tl send and T l receive All are slotted time values.

[0020] At the same time, the data packet transmission success rate on the i-th link is calculated according to the following formula:

[0021]

[0022] Where max(send_count) is the maximum number of received encoded data packets in the same batch.

[0023] In step 3, if the arrival rate of the data packets at the sending end is much greater than the service rate and the buffer area is long enough, the main factors affecting the transmission delay are channel delay, service rate and traffic scheduling strategy. When optimizing the service scheduling ratio, there is no need to consider the queuing delay of the data packets at the sending end. The calculation steps of the optimal scheduling ratio are as follows:

[0024] (3-1a) Determine the shortest transmission delay T opt When all channels complete the transmission task at the same time, the overall transmission delay will reach the minimum value, that is, the channel utilization rate is maximized. Consider m channels (without loss of generality, assuming that the propagation delay d1≥d2≥…d m-1 ≥d m ), the expected value of the shortest number of transmission time slots required for n services to complete one decoding is:

[0025]

[0026] where d i and r i Determined by formulas (1) and (2) in step (2-2) respectively.

[0027] (3-2a) Determine the optimal scheduling ratio ρ of each service flow j The SDN controller calculates the proportion of scheduled services for the jth service on each channel according to the principle of service transmission fairness.

[0028]

[0029] That is, all channels are scheduled for traffic according to the above ratio.

[0030] In step 3, if the service flow arrival rate at the sending end is not much greater than the service rate, the packet queuing delay factor needs to be considered when calculating the optimal scheduling ratio. At this time, the calculation steps of the shortest transmission delay and the optimal scheduling ratio of each service are as follows:

[0031] (3-1b) Determine the shortest transmission delay T optWhen all channels complete the transmission task at the same time, the overall transmission delay will reach the minimum value, that is, the channel utilization rate is maximized. Consider m channels (without loss of generality, assuming that the propagation delay d1≥d2≥…d m-1 ≥d m ), the expected value T of the shortest number of transmission time slots required for n services to complete one decoding opt Satisfies the following equation

[0032]

[0033] where q j Complete a batch of N for the jth service flow at the sender j The average queuing time required for a data packet is d i and r i Determined by formulas (1) and (2) in step (2-2) respectively. Note that the left side of formula (5) is about T opt Monotonic function, you can set T opt Initial value T opt =max{q j +d i}, and then use the step method to update T opt =T opt +δ(δ>0) makes equation (5) equal, and the solution is T opt .

[0034] (3-2b) Determine the optimal scheduling ratio ρ of each service flow j The SDN controller calculates the proportion of scheduled services for the jth service on each channel according to the principle of service transmission fairness.

[0035]

[0036] Where T opt Determined by step (3-1b).

[0037] Compared with the prior art, the technical effects of the present invention are as follows:

[0038] The present invention utilizes the excellent property that fountain code decoding is independent of the order in which data packets arrive. In a transmission system based on fountain codes, the receiving end only needs to receive a certain number of encoded data packets to decode the original data. Therefore, the optimization of system transmission performance only needs to consider the two goals of maximizing throughput (minimizing transmission delay) and business service fairness, without considering the order in which data packets arrive. Compared with other efforts to optimize transmission performance under the condition of in-order delivery of data packets in traditional transmission systems, the scheduling strategy proposed in the present invention greatly reduces the complexity of the model and the uncertainty of data packet disorder, and has better versatility.

[0039] The encoding and decoding process of a batch of data packets in a multi-connection concurrent transmission system based on Raptor code is taken as the research object. The business traffic optimization model is established by combining relevant variables such as multi-business traffic information and multi-channel state information, and the traffic scheduling ratio of each channel in the multi-business multi-channel scenario is optimized and calculated. The proposed scheduling optimization strategy is simulated in the MATLAB environment, and the results show that the expected effect of multi-business service fairness and minimum overall transmission delay is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a system block diagram and scheduling transmission schematic diagram of a multi-service multi-path concurrent transmission system based on fountain codes.

[0041] Figure 2 It is a flow chart of the multi-service multi-path concurrent transmission traffic scheduling method based on fountain code of the present invention.

[0042] Figure 3 This is the result graph of the average service transmission delay (without queuing delay).

[0043] Figure 4 This is the result graph of the average service transmission delay (including queuing delay). DETAILED DESCRIPTION

[0044] The present invention is further explained below in conjunction with the accompanying drawings and embodiments, but the protection scope of the present invention should not be limited thereto.

[0045] The invention relates to network devices such as software defined network (SDN) controllers, terminal nodes (transmitters, receivers), network nodes (switches, routers), etc.

[0046] In such Figure 1 In the multi-connection concurrent transmission system based on fountain codes shown in the figure, both the sender and the receiver have multiple network interfaces at the same time, and the data flow between the two can be split into multiple sub-data flows for transmission on different paths; it is assumed that the SDN controller has configured multiple transmission paths between the sender and the receiver. On this basis, the SDN controller controls the sender to forward the data packets to multiple network interfaces for transmission at the optimal scheduling ratio according to the periodically obtained link status and sender load status information.

[0047] By utilizing the implementation method of the present invention, a multi-connection concurrent transmission system can achieve minimum transmission delay while ensuring service fairness among different business flows.

[0048] like Figure 2 As shown, the specific implementation process of the invention is divided into 7 steps:

[0049] Step 1. The sender obtains the number of original data packets of the service flow received locally and encodes them, and reports the encoding length of each service flow to the SDN controller. The specific steps are as follows:

[0050] (1-1) The sender periodically obtains the relevant status information of 3 (n=3) local service flows, including the number K of original data packets of each service in the current cache queue. j ,j=1,...,3 (i.e., the length of the cache queues of multiple services), the waiting time of each service q j ;

[0051] (1-2) The sender determines the encoding length N of each service flow according to the original data packet length of each service flow and the RaptorQ standard. j , encode the original data packet. In this embodiment, N corresponding to the three services j Randomly select from N1 = 10, 12; N2 = 18, 20; N3 = 26, 30. The sender sets the encoding length N of each service flow j Report to the SDN controller.

[0052] Step 2. The receiving end periodically obtains the status information of a total of 3 (m=3) channels, including the propagation delay d of channel i (i=1,...,3) i , transmission success rate r i (the probability of successfully transmitting a data packet within a unit time slot), etc., and report to the SDN controller. The specific steps are as follows:

[0053] (2-1) Before forwarding the coded data packet to the i-th (i=1,...,3) channel, the sender writes fields for channel measurement in the header of each coded data packet, including: a). link number link_ID to which the data packet is forwarded; b). send batch number Send_batch_ID; c). send time T_send; d). send number send_count, which is reset to 1 when a new batch of coded data packets starts to be sent.

[0054] (2-2) The receiving end counts the received coded data packets, parses the corresponding fields in the packet header such as the data packet batch number, sending time and sending number, and calculates the average delay of the i-th link according to the following formula:

[0055]

[0056] In the above formula, sum_receive is the total number of packets received in the same batch, T l receive is the time when the lth data packet arrives at the receiving end, T l send and Tl receive All are slotted time values.

[0057] At the same time, the data packet transmission success rate on the i-th link is calculated according to the following formula:

[0058]

[0059] Where max(send_count) is the maximum number of received encoded data packets in the same batch.

[0060] Step 3. The SDN controller calculates the optimal service scheduling ratio for multi-channel concurrent transmission based on the service flow information fed back by the sender and the channel status information fed back by the receiver, in accordance with the service fairness principle and the principle of maximizing channel utilization. The specific steps are as follows:

[0061] If the arrival rate of the data packets at the sender is much greater than the service rate and the buffer area is long enough, the main factors affecting the transmission delay are channel delay, service rate and traffic scheduling strategy. When optimizing the service scheduling ratio, there is no need to consider the queuing delay of the data packets at the sender. The calculation steps of the optimal scheduling ratio are as follows:

[0062] (3-1a) Determine the shortest transmission delay T opt When all channels complete the transmission task at the same time, the overall transmission delay will reach the minimum value, that is, the channel utilization rate is maximized. Consider 3 (m = 3) channels (without loss of generality, assuming that the propagation delay d1 ≥ d2 ≥ ... d m-1 ≥d m ), the expected value of the shortest number of transmission time slots required for n services to complete one decoding is:

[0063]

[0064] where d i and r i Determined by formulas (1) and (2) in step (2-2) respectively.

[0065] (3-2a) Determine the optimal scheduling ratio ρ of each service flow j The SDN controller calculates the proportion of scheduled services for the jth service on each channel according to the principle of service transmission fairness.

[0066]

[0067] That is, all channels are scheduled for traffic according to the above ratio.

[0068] If the service flow arrival rate at the sender is not much greater than the service rate, the packet queuing delay factor needs to be considered when calculating the optimal scheduling ratio. In this case, the calculation steps for the shortest transmission delay and the optimal scheduling ratio of each service are as follows:

[0069] (3-1b) Determine the shortest transmission delay T opt When all channels complete the transmission task at the same time, the overall transmission delay will reach the minimum value, that is, the channel utilization rate will be maximized. Considering 3 channels, the expected value of the shortest number of transmission time slots required for 3 services to complete a decoding is T opt Satisfies the following equation

[0070]

[0071] where q j Complete a batch of N for the jth service flow at the sender j The average queuing time required for a data packet is d i and r i Determined by formulas (1) and (2) in step (2-2) respectively. Note that the left side of formula (5) is about T opt Monotonic function, you can set T opt Initial value T opt =max{q j +d i}, and then use the step method to update T opt =T opt +δ(δ>0) makes equation (5) equal, and the solution is T opt .

[0072] (3-2b) Determine the optimal scheduling ratio ρ of each service flow j The SDN controller calculates the proportion of scheduled services for the jth service on each channel according to the principle of service transmission fairness.

[0073]

[0074] Where T opt Determined by step (3-1b).

[0075] Step 4: The SDN controller sends the optimal service scheduling ratio to the sending end.

[0076] Step 5. The sender dispatches the service according to the service scheduling ratio ρ issued by the SDN controller j , j = 1, ..., 3, schedule the encoded data packet to the corresponding channel (network interface) and send it.

[0077] Step 6. The receiving end receives the coded data packets of each service stream transmitted via 3 (m=3) channels. When receiving the N packets of the jth service, j After a data packet is decoded, the K corresponding to the service is obtained. j The original data packet is sent to the sender, and the ACK corresponding to service j is sent to the sender j .

[0078] Step 7. The sender detects the received ACK status of each service flow. j After (j=1,...,3), the transmission ends.

[0079] Figure 3 and Figure 4 The results of the average transmission delay of the services without queuing delay and the average transmission delay of the services with queuing delay are shown respectively. The results show that the average transmission delay of multiple service flows is basically the same, and this result still holds true when the coding length of each service flow changes, indicating that the scheduling strategy of the present invention can ensure the fairness of service. At the same time, the average transmission delay of the service changes with the channel parameters, indicating that the scheduling strategy of the present invention can dynamically adjust the scheduling strategy according to the channel parameters.

Claims

1. A traffic scheduling method for a multi-path concurrent transmission system based on fountain codes, the system comprising a transmitter and a receiver having multiple network interfaces at the same time, and the data stream between the transmitter and the receiver can be split into multiple sub-data streams for transmission on different paths, and an SDN controller configures multiple transmission paths between the transmitter and the receiver, characterized in that: The method comprises the following steps: Phase 1: Sending side: Periodically obtain the local n service flow status information, including the number K of original data packets of each service in the current cache queue j , j = 1, ..., 3, the waiting time q of each service j ; Determine the encoding length N of each service flow j Report to the SDN controller; Encode the original data packets of each service flow; Before forwarding the coded data packet to the i-th channel, writing a field for channel measurement into the header of the coded data packet; Receiver: Periodically obtain m channel status information, including the propagation delay d of channel i i , transmission success rate r i , i = 1, ..., m, and report to the SDN controller; Count the total number of coded packets received, parse the fields used for channel measurement in the header of each coded packet, and calculate the average delay d of the i-th link i And the packet transmission success rate r i , the formula is as follows: Where sum_receive is the total number of coded packets received in the same batch, T l sen d is the time when the first coded data packet is sent from the sender, T l receive is the time when the first encoded data packet arrives at the receiving end, T l send and T l receive All are time values ​​after slotting; max(send_count) is the maximum number of received coded data packets in the same batch; Phase 2: SDN Controller: According to the service flow status information fed back by the transmitter and the channel status information fed back by the receiver, the optimal scheduling ratio ρ of the jth service is calculated j , j = 1, ..., n, and sent to the sending end; Sending side: According to the optimal scheduling ratio ρ issued by the SDN controller j , j = 1, ..., n, schedule the coded data packet to the corresponding channel and send it; Receiver: Receive the coded data packets of each service stream transmitted via m channels. When receiving N packets of the jth service j After a data packet is decoded, the K corresponding to the service is obtained. j The jth service receives the original data packet and sends the encoded data packet ACK corresponding to the jth service to the sender. j ; Phase 3: Sending side: Detect the ACK status of the coded data packets received for each service flow. When the ACK status of the coded data packets of all services is received, j After that, end this transmission; Calculate the optimal scheduling ratio ρ of the jth service j , as follows: When the arrival rate of the data packets at the sender is much greater than the service rate and the buffer is long enough, the optimal scheduling ratio ρ of the jth service is calculated j The steps include: (3-1a) Determine the shortest transmission delay T opt :When all channels complete the transmission task at the same time, the overall transmission delay will reach the minimum value, that is, maximize the channel utilization. Consider m channels and assume that the propagation delay d1≥d2≥…d m-1 ≥d m , the expected value of the shortest number of transmission time slots required for n services to complete one decoding is: (3-2a) Determine the optimal scheduling ratio ρ of each service flow j : The SDN controller calculates the proportion of scheduled services for the jth service on each channel according to the principle of service transmission fairness: That is, all channels are scheduled for traffic according to the above ratio; When the service flow arrival rate at the sender is not much greater than the service rate, the optimal scheduling ratio ρ of the jth service is calculated j , including the following steps: (3-1b) Determine the shortest transmission delay T opt :When all channels complete the transmission task at the same time, the overall transmission delay will reach the minimum value, that is, maximize the channel utilization. Considering m channels, the propagation delay d1≥d2≥…d m-1 ≥d m , the expected value T of the shortest number of transmission time slots required for n services to complete one decoding opt Satisfies the following equation where q j Complete a batch of N for the jth service flow at the sender j The average queuing time required for a data packet is the left side of formula (5) about T opt Monotonic function, you can set T opt Initial value T opt =max{q j +d i }, and then use the step method to update T opt =T opt +δ, so that equation (5) takes an equal sign, where δ>0, and the solution is T opt ; (3-2b) Determine the optimal scheduling ratio ρ of each service flow j :The SDN controller calculates the proportion of scheduled services for the jth service on each channel according to the principle of service transmission fairness 2. The traffic scheduling method for a multipath concurrent transmission system based on fountain codes according to claim 1 is characterized in that: The fields used for channel measurement include: link number link_ID to which the data packet is forwarded, sending batch number Send_batch_ID, sending time T_send, sending number send_count, which is reset to 1 when starting to send a new batch of encoded data packets.

3. The traffic scheduling method for a multipath concurrent transmission system based on fountain codes according to claim 1 is characterized in that: The fields used for channel measurement include: link number link_ID to which the data packet is forwarded, sending batch number Send_batch_ID, sending time T_send and / or sending number send_count.

Images