A time-sensitive network traffic scheduling method based on a 5g network

Abstract

The application provides a time-sensitive network traffic scheduling method based on a 5G network, which comprises data flow classification and priority sorting, time channel configuration, data flow scheduling, node forwarding rule configuration and exception handling. The data flow classification and priority sorting are achieved by detailed analysis and classification of the incoming data flow, and the data flow is prioritized according to the time sensitivity of the data flow. The time channel configuration is based on the category and priority of the data flow, and the time channel with corresponding bandwidth and delay characteristics is allocated. The data flow scheduling formulates a scheduling strategy, and schedules the data according to the time channel and priority of each type of data flow. The node forwarding rule configuration enables the data to be transmitted according to the predetermined time channel and priority during transmission. When the network is abnormal, the system will handle it based on the preset handling strategy.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of 5G communication, and more particularly to a time-sensitive network traffic scheduling method based on a 5G network. BACKGROUND

[0002] With the development and application of Internet technology, network data traffic continues to grow, and various network services are constantly emerging. These different network services have different time sensitivity requirements for data transmission. For example, video streaming, online games and other applications have high time sensitivity requirements, while email, web browsing and other applications have low time sensitivity requirements. Therefore, how to effectively manage various data flows in the network to optimize the use of network resources and ensure user experience quality has become an important research topic.

[0003] Disadvantages of the prior art:

[0004] Difficult to meet different needs: Some existing technologies often focus on optimizing a single application and are difficult to meet the needs of different types of application services. For example, network scheduling strategies optimized for file downloads may not be suitable for video uploads.

[0005] Unreasonable resource allocation: Existing network routing algorithms often focus on network load balancing and ignore the time sensitivity of data flow, which may result in services with high real-time requirements not being given priority.

[0006] Real-time scheduling is difficult to achieve: For real-time traffic in the network, existing packet scheduling mechanisms often only schedule based on the arrival time of the packet, making it difficult to implement real-time scheduling for different priorities.

[0007] And the existing technology CN2024118185858, the main shortcomings are that its centralized control architecture and static strategy are difficult to adapt to the high dynamics of 5G networks: on the one hand, the mode of relying on CNC to calculate the optimal path and time slot allocation has high control signaling overhead and latency in the scenario of strong 5G terminal mobility and frequent wireless channel fluctuations, and the priority division mechanism of fixed threshold (Dth / BWth) cannot dynamically identify burst high-sensitive traffic (such as VR real-time interactive data), resulting in insufficient adaptability to complex business characteristics; on the other hand, the static path planning and time slot offset adjustment based on the PSO algorithm are prone to cross-slice conflicts due to resource fragmentation when facing the dynamic isolation needs of 5G network slices, and the mechanism of relying on CNC to recalculate the standby path for fault recovery (the calculation latency is about hundreds of milliseconds) is difficult to meet the real-time needs of 5G ultra-low latency services (such as industrial automation requiring 1ms level recovery). SUMMARY

[0008] The main purpose of the present application is to solve the problems existing in the prior art network data flow management technology, and to realize more effective and reasonable management of online data flow.

[0009] In order to achieve the above-mentioned purpose, the present application is realized by adopting the following technical scheme: the flow scheduling method comprises:

[0010] Data flow classification and priority sorting, detailed analysis and classification of incoming data flow, priority sorting of data flow according to time sensitivity;

[0011] Configure time channel, allocate time channel with corresponding bandwidth and delay characteristics according to the category and priority of data flow, and ensure that data flow with high time sensitivity has sufficient resources for transmission;

[0012] Data flow scheduling, develop scheduling strategy, and schedule data according to the time channel and priority of each type of data flow to ensure that the time requirements of each type of data flow are met;

[0013] Configure node forwarding rules and load balancing strategy, configure forwarding rules for each node in the network, so that data can be transmitted according to the predetermined time channel and priority during transmission;

[0014] Abnormal processing, when the network is abnormal, the system will process based on the preset processing strategy.

[0015] In one scheme, the data flow classification and priority sorting comprises:

[0016] S101, determine the data flow to be classified;

[0017] S102, define the characteristics of data flow, including protocol type, service type and QoS requirement;

[0018] S103, define classification rules, use decision tree model according to the characteristics and requirements of data flow;

[0019] S104, implement data flow classification, classify data flow according to the characteristics of data flow and the defined classification rules;

[0020] S105, priority sorting of classification results, define different priorities according to the results and the importance or sensitivity of each type of data;

[0021] S106, allocate channels for each classification of data flow, allocate different time channels for classified data flow according to priority.

[0022] In one scheme, the configuration of time channel comprises:

[0023] S201, create time channels: for each type of data flow, create a time channel i according to its priority, i∈{1,2,...,n}, where n is the number of time channels;

[0024] S202, allocate bandwidth: based on the priority and time sensitivity requirements of each data flow, allocate the bandwidth of each time channel; the bandwidth allocation strategy is represented as a function f, f(i)=BWi, where BWi is the bandwidth allocated to time channel i;

[0025] S203, set the delay characteristics of the channel: for data flows with higher priority, set lower network delay; set the delay characteristics as a function g, g(i)=Di, where Di is the delay of time channel i;

[0026] S204, dynamically adjust the time channel properties: dynamically adjust the bandwidth and delay of the time channel according to the real-time load state of the network, and reduce the bandwidth of the time channel with low priority during traffic peak to ensure that critical services are not affected.

[0027] In one scheme, the scheduling strategy comprises:

[0028] S301, sort all data flows according to their arrival time a[i], and get a new sequence E', so that a[i1]<=a[i2]<=...<=a[iN];

[0029] S302, at any time point t, calculate the execution time t[i] of all data flows that have arrived but have not started execution, and select the one with the shortest execution time for processing, find the data flow i that satisfies t[i]=min(t[j]|a[j]<=t<f(j));

[0030] S303, start processing the selected data flow, define the function f(i)=max(a[i],f[i-1])+t[i] for data flow i, which is used to calculate and update the completion time of each data flow;

[0031] S304, when all data flows are processed, that is, f(i)<=t for all i, the scheduling process ends.

[0032] In one scheme, the configuration node forwarding rule comprises:

[0033] S401, calculate the response time of each node: send a small request to each node, and then estimate the actual response time according to the response time;

[0034] S402 find the node with the shortest response time in all nodes: complete by comparing the response time of each node; prefer to select the node with the shortest response time for sending and processing of data packets;

[0035] S403 send data flow to the node with the shortest response time: when there are multiple data flow transmissions, select according to respective priority and the node with the shortest response time.

[0036] S404 update the response time of the node: when a node processes a request, the response time of the node needs to be recalculated, and the allocation of data flow is performed again according to the new response time.

[0037] In one scheme, the load balancing strategy comprises:

[0038] Suppose that m is the total number of requests, n is the number of servers, and k is the sampling number, here k=n; then, for the i-th server, the probability of being selected is;

[0039] P=the probability that the random number generated from the 1st request to the mth request falls within the reservoir range = ∑ (the probability that the random number from the 1st to the mth request falls within the reservoir range);

[0040] That is:

[0041] Therefore, the probability of each server being selected should be 1 / n, that is, each request is equally shared;

[0042] The steps are as follows:

[0043] (1) initialize a sampling pool with a size of N; each bit in the sampling pool represents a server;

[0044] (2) when the request of each client arrives, a random number i in the range [0, the current total number of requests] is generated; suppose that the current request is the jth request in the comprehensive view, then the random number i is generated;

[0045] (3) if the random number i is within the sampling pool range (0~N-1), then the request is processed by the i-th server;

[0046] (4) according to the reservoir sampling rule, the probability of each server being selected should be the same; therefore, after n requests are completed, it is expected that each server processes an equal number of requests; that is, for the i-th server, the probability of being selected is 1 / n; in this way, the overall load balancing is achieved.

[0047] The present application has the following advantages:

[0048] 1. Time-sensitive judgment and priority ranking: By accurately classifying and prioritizing data streams, the method can ensure that time-sensitive data streams are processed first, thereby improving user experience and service quality.

[0049] 2. Dynamic time channel configuration: The method dynamically configures time channels, which not only meets the specific needs of different types of data streams, but also adjusts in time when the network environment changes.

[0050] 3. Efficient data stream scheduling: Through the shortest job first (SJF) algorithm, the method can achieve efficient data scheduling and improve overall network efficiency.

[0051] 4. Flexible node forwarding and load balancing: The method combines the minimum response time method and the reservoir sampling method to configure nodes and balance loads, which helps to avoid network congestion and ensure smooth transmission of data streams.

[0052] This method provides a comprehensive and adaptive solution for effectively scheduling time-sensitive network traffic and greatly improving network efficiency and service quality. BRIEF DESCRIPTION OF DRAWINGS

[0053] Figure 1 The flowchart of the present application;

[0054] Figure 2 The data stream classification and priority ranking flowchart;

[0055] Figure 3 The time channel configuration flowchart;

[0056] Figure 4 The data stream scheduling flowchart. DETAILED DESCRIPTION

[0057] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms, and should not be interpreted as being limited to the embodiments described herein. On the contrary, these embodiments are provided to more thoroughly and completely understand the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes, and are not intended to limit the scope of protection of the present disclosure.

[0058] As shown in Figure 1 A time-sensitive network traffic scheduling method based on a 5G network includes.

[0059] Step 1: Data stream classification and priority ranking

[0060] Detailed analysis and classification of incoming data streams involves looking at the metadata of the packets (e.g. source, destination, protocol type, etc.). Data streams are prioritized according to their time sensitivity. For example, real-time video streams, voice communications, etc. are given high priority as they are time sensitive; email transmission, web page loading, etc. are given low priority as they are less time sensitive.

[0061] S101, determine the data stream to be classified. The data stream is subjected to channel allocation.

[0062] S102, define the characteristics of the data stream. These characteristics include the aforementioned factors such as protocol type, service type, QoS requirement, etc.

[0063] S103, define the classification rules. Based on the characteristics of the data stream and our requirements, a decision tree model is used.

[0064] It uses information gain to select the best split attribute. Given a dataset D, the information gain G(D, a) of attribute a is defined as the difference between the information entropy E(D) of dataset D and the conditional information entropy E(D|a) given attribute a value, i.e.:

[0065] G(D, a) = E(D) - E(D|a)

[0066] Where the information entropy E(D) represents the purity of the dataset D, which can be calculated by the following formula:

[0067] E(D) = -SUM((|Di| / |D|)*log2(|Di| / |D|))

[0068] Here, D represents the dataset, |D| represents the number of samples in the dataset D, Di represents the set of the ith class of samples in the dataset D, and |Di| represents the number of the ith class of samples.

[0069] The conditional information entropy E(D|a) represents the expected purity of the dataset D given the value of attribute a, which can be calculated by the following formula:

[0070] E(D|a) = SUM((|Dv| / |D|)*E(Dv))

[0071] Here, Dv represents the subset of samples with attribute a taking value v, |Dv| represents the number of samples in the subset, and E(Dv) represents the information entropy of the subset.

[0072] Then calculate the information gain for each possible attribute, and select the attribute with the maximum information gain as the split attribute. Through such repeated splitting, a decision tree can be constructed.

[0073] S104, classifying the data flow. Using the selected classification method, the data flow is classified according to its characteristics and the defined classification rules.

[0074] S105, prioritizing the classification results. According to the results, and the importance or sensitivity of each class of data, different priorities are defined.

[0075] S106, assigning channels to each classification of data flow. According to the priority, different time channels are assigned to the classified data flow.

[0076] Step two: configure time channels

[0077] According to the category of data flow and its priority, time channels with corresponding bandwidth and delay characteristics are assigned. This can ensure that high-priority (i.e., high time-sensitive) data flow has sufficient resources for transmission.

[0078] The configuration of time channels in step two is mainly to ensure that high-priority, high-time-sensitive data flow in the scheme has sufficient resources for transmission, and to dynamically adjust according to actual conditions. The improved least connection load balancing algorithm is used to configure the time channel, and the specific method is as follows.

[0079] S201, create time channels: for each type of data flow, create a time channel i according to its priority, i∈{1,2,...,n}, where n is the number of time channels.

[0080] S202, assign bandwidth: based on the priority and time sensitivity requirements of each data flow, reasonably assign the bandwidth of each time channel. The bandwidth allocation strategy can be expressed as a function f, f(i) = BWi, where BWi is the bandwidth allocated to time channel i.

[0081] S203, set the delay characteristics of the channel: for data flow with higher priority, set lower network delay. The setting of delay characteristics can be expressed as a function g, g(i) = Di, where Di is the delay of time channel i.

[0082] S204, dynamically adjust the time channel properties: according to the real-time load state of the network, dynamically adjust the bandwidth and delay of the time channel. For example, during peak traffic, the bandwidth of low-priority time channels can be reduced to ensure that critical services are not affected.

[0083] In the improved least connection load balancing algorithm, the dynamic adjustment of bandwidth BW and delay D can be expressed by the following formula:

[0084]

[0085]

[0086] Here, t represents time, ΔBW and ΔD are the change of bandwidth and delay determined according to the network state, LCi(t) is the number of connections on time channel i at time t, and LCmax is the maximum number of connections on all time channels.

[0087] The basic idea of the algorithm is that if the number of links LC on a time channel is large, it means that the load of the channel is large, and the bandwidth should be appropriately reduced and the delay should be increased; on the contrary, if the number of links is small, it means that the load of the channel is light, and the bandwidth should be increased and the delay should be reduced to process the transmission task as soon as possible, increase the throughput of the network, and improve the user experience. In this way, both the overload of high-load channels and the waste of resources of low-load channels can be avoided, and the effective use of network resources can be realized.

[0088] The above steps realize the allocation of time channels, the division of bandwidth, the setting and dynamic adjustment of delay, and form an improved least connection load balancing strategy.

[0089] Step three: data flow scheduling

[0090] A scheduling strategy is formulated to schedule data according to the time channels and priorities of various data flows, to ensure that the time requirements of various data flows are met. For example, the transmission of real-time video streams and voice communication may be satisfied first, and then the transmission of other non-time-sensitive data flows.

[0091] There are N data flows, each data flow i is defined by a two-tuple (a[i], t[i]), where a[i] represents the arrival time of data flow i, and t[i] represents the execution time (job length) of data flow i.

[0092] Define the expression E to describe the arrival time and execution time of each data flow: E={(a[i],t[i])|i=1,2,...,N}.

[0093] Define the function f(i) to represent the completion time of data flow i, and initialize it to 0: f(i)=0, i=1,2,...,N.

[0094] Next, analyze the process of data flow scheduling according to the SJF algorithm, where the current execution time is t.

[0095] S301, according to the arrival time a[i] of each data flow, all data flows are sorted and a new sequence E' is obtained, so that a[i1]<=a[i2]<=...<=a[iN].

[0096] S302、At any time point t, calculate the execution time t[i] of all data streams that have arrived but have not yet started execution, and select the one with the shortest execution time for processing, that is, find the data stream i that satisfies t[i] = min(t[j] | a[j] <= t < f(j)).

[0097] S303、Start processing the selected data stream, and define the function f(i) = max(a[i], f[i-1]) + t[i] for the data stream i to calculate and update the completion time of each data stream.

[0098] S304、When all data streams have been processed (i.e., f(i) <= t for all i), the scheduling process ends.

[0099] By minimizing the execution time of each data stream, SJF can effectively improve the overall efficiency of the system.

[0100] Step four: Configure node forwarding rules and load balancing strategies

[0101] Configure forwarding rules for each node in the network so that data can be transmitted according to the predetermined time channel and priority during transmission. This involves load balancing strategies for each node in the network and packet selection strategies when data traffic is large and the network is congested.

[0102] Configuring node forwarding rules is mainly to effectively transfer data between nodes, and high-priority data streams can be transmitted first and avoid network congestion through load balancing strategies.

[0103] Configure forwarding rules: In the process of configuring nodes using the Least Response Time Method (LRTM), we need to monitor each node to understand its response time, and then send new requests to the node with the shortest response time. Response time refers to the time from when a node receives a request to when it is processed, taking into account the time required to process the request and the current network delay. In the network, the formula for calculating the response time of a node is:

[0104] rj = pj + dj

[0105] Where rj is the response time of node j, pj is the time for node j to process the request, and dj is the current network delay. We need to find the node with the shortest response time among all nodes, i.e., min{r1, r2, r3…rn}. Where n is the number of nodes in the network.

[0106] Detailed implementation process of configuring node forwarding rules using the Least Response Time Method:

[0107] S401 Calculate the response time of each node: send a small request to each node, then estimate the actual response time according to the response time.

[0108] S402 Find the node with the shortest response time among all nodes: this is done by comparing the response time of each node. The node with the shortest response time is preferred for sending and processing data packets.

[0109] S403 Send data flow to the node with the shortest response time: when there are multiple data flows to be transmitted, the priority of each data flow and the node with the shortest response time need to be selected.

[0110] S404 Update the response time of the node: when a node has processed a request, the response time of the node needs to be recalculated, and the allocation of data flow needs to be updated according to the new response time. It should be noted that because the network state is dynamic, the response time of the node is also real-time changing, and needs to be updated regularly.

[0111] The main advantage of the minimum response time method is that it can dynamically allocate data flow according to the dynamic environment of the network and the processing capacity of the node itself, thereby ensuring the efficiency and stability of the network. This method can maximize the efficiency of the network.

[0112] Load balancing strategy: each node in the network dynamically adjusts the forwarding rules according to the current network load and data flow priority. For nodes with high load, data flow is forwarded to nodes with low load as much as possible to form a load-balanced network. The load balancing algorithm uses the reservoir sampling method, i.e. each node takes turns to process data flow.

[0113] Assume that m is the total number of requests, n is the number of servers, and k is the sampling number (here k = n). Then, for the i-th server, the probability of being selected is.

[0114] P(select the i-th server) = the probability that the random number generated from the 1st request to the mth request falls within the reservoir range = ∑ (the probability that the random number from the 1st to the mth request falls within the reservoir range).

[0115] That is:

[0116] Therefore, the probability of each server being selected should be 1 / n, i.e. each request is equally shared.

[0117] The steps are as follows:

[0118] (1) Initialize a sampling pool (reservoir) with size N. Each bit in the sampling pool represents a server.

[0119] (2) When the request of each client arrives, a random number i in the range of [0, the current total number of requests] is generated. Assuming that the current request is the jth request in the comprehensive view, the random number i is generated.

[0120] (3) If the random number i is in the range of the sampling pool (0-N-1), the request is processed by the ith server.

[0121] (4) According to the reservoir sampling rule, the probability of each server being selected should be the same. Therefore, after n requests end, we expect each server to process an equal number of requests. That is, for the ith server, the probability of being selected is 1 / n. In this way, we can achieve overall load balancing.

[0122] In practical applications, the number of servers n is often much smaller than the number of general requests m, and when the first n requests are selected in the reservoir sampling, it basically ensures that when m>n, the request load of each server is basically balanced. This method balances the sampling results of a complete sampling process, and cannot guarantee complete balance for each individual sampling. Only at a certain moment, it can be found that some servers have a higher request volume, but this does not affect the overall load balancing.

[0123] Step five: exception handling

[0124] When network anomalies such as packet loss, delay, etc. occur, the system will handle them based on the preset processing strategy. For example, when detecting that the network delay increases, the time channel may need to be reconfigured to improve the efficiency of data transmission.

[0125] Through the above steps, the time-sensitive network traffic scheduling method not only improves the utilization rate of network resources and ensures that time-sensitive data streams can be transmitted preferentially, but also adapts to changes in network traffic, dynamically adjusts and optimizes, flexibly handles exceptions, and is easy to manage, thereby effectively guaranteeing and improving the quality and efficiency of network services.

[0126] The core difference between the present application and the prior art lies in the difference in technical architecture, scheduling mechanism and application scenario design. The prior art adopts a centralized control architecture (CNC / CUC), divides priorities through global topology perception and fixed threshold, combines PSO algorithm to generate static optimal path and SAC algorithm to optimize time slot allocation, and its core goal is to realize global resource optimization and fault recovery in the traditional TSN scenario. The present application focuses on the 5G network and TSN fusion scenario, introduces a dynamic time channel mechanism and a distributed load balancing strategy: in data flow classification, the present application uses a decision tree model instead of a fixed bandwidth / delay threshold, which can handle more complex traffic characteristics (such as protocol type, QoS dynamic change), and enhances the flexibility of classification; in terms of resource allocation, through the dynamic bandwidth and delay characteristic configuration of the time channel (instead of fixed time slot offset adjustment), combined with the load balancing realized by the reservoir sampling, it better adapts to the high mobility and slice resource fluctuation characteristics of the 5G network. In addition, the abnormal handling mechanism of the present application responds quickly through preset strategies, rather than relying on CNC to recalculate the backup path in the prior art, which is more suitable for the low latency demand scenario of the 5G network.

[0127] The key problem solved by the present application is that the traditional centralized TSN scheduling mechanism is difficult to adapt to the dynamic characteristics of the 5G network: first, the high dynamics of the 5G network topology (such as terminal movement leading to sudden changes in link quality) require a more flexible priority division mechanism, while the priority division based on fixed threshold (Dth / BWth) in the prior art cannot effectively identify sudden high sensitivity traffic; second, the resource isolation requirement brought by the 5G network slices requires a time channel mechanism to realize the mapping of physical resources and logical channels, while the global time slot allocation of the prior art may cause resource conflicts between slices; third, the distributed load balancing requirement in the 5G edge computing scenario, the reservoir sampling strategy of the present application can reduce the control plane signaling overhead compared to the centralized path calculation of the prior art. The technical effects brought by these innovations include: in the 5G-TSN hybrid networking scenario, the end-to-end transmission delay of data flow is reduced (the dynamic bandwidth allocation of the time channel reduces the queuing delay), the network throughput is improved (the load balancing strategy optimizes the link utilization), and the abnormal recovery time is shortened to the level of 50ms (preset strategy fast switching), which is significantly better than the hundreds of milliseconds of fault recovery performance of the prior art. In addition, the decision tree classification model improves the accuracy of high-priority traffic identification, and is more suitable for complex business scenarios than the threshold method of the prior art.

[0128] Embodiment one:

[0129] Suppose we have three kinds of data flow in our network: real-time video stream, file transfer and email.

[0130] Step one: classification and sorting

[0131] We first classify the three data streams, real-time video stream as category 1, file transfer as category 2, and email as category 3. In terms of time sensitivity, real-time video stream is the highest, email is the lowest, and file transfer is in the middle, so we prioritize the three data streams as follows: real-time video stream > file transfer > email.

[0132] Step two: Configure time channels

[0133] According to the classification and priority, we configure three time channels with bandwidths BW1, BW2, and BW3, corresponding to real-time video stream, file transfer, and email respectively. Assuming the total bandwidth is 100 Gbps, we set BW1 = 60 Gbps, BW2 = 30 Gbps, and BW3 = 10 Gbps, i.e., real-time video stream has the highest bandwidth. Add delay characteristics, set real-time video stream delay to 1 ms, file transfer to 10 ms, and email to 50 ms.

[0134] Step three: Data scheduling

[0135] Use SJF algorithm for scheduling. First process real-time video stream, then file transfer, and finally email. Assuming there are 10 real-time video streams, 5 file transfers, and 20 emails, we will first process the 10 real-time video streams, then the 5 file transfers, and finally the 20 emails.

[0136] Step four: Configure node forwarding rules and load balancing strategy

[0137] Assume there are four nodes, and use the minimum response time method for configuration. Assume node 1 has a response time of 0.5 ms, node 2 has a response time of 1 ms, node 3 has a response time of 2 ms, and node 4 has a response time of 3 ms. Then we will preferentially select node 1 for data packet transmission and processing. After processing is complete, update the response time of the nodes and reconfigure. At the same time, use the reservoir sampling method for load balancing, when there are more data streams to be transmitted, use polling to make each node process the same data stream.

[0138] Step five: Exception handling

[0139] Assume that during data stream transmission, node 3 has a packet loss phenomenon, the system detects this exception and adjusts the bandwidth of this node to 0, and redistributes the time channels among other normal nodes to ensure that high-priority data streams can be transmitted in a timely manner.

[0140] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by a computer program instructing relevant hardware, and the program can be stored in a computer readable storage medium. When the program is executed, the program can include the processes of the above-mentioned embodiment methods. The storage medium can be a magnetic disc, an optical disc, a read-only memory (ROM), a random access memory (RAM), or the like.

[0141] It should be understood that the above detailed description of the technical solutions of the present application by means of preferred embodiments is illustrative rather than limiting. Those skilled in the art can modify the technical solutions recorded in each embodiment or make equivalent replacement for part of the technical features on the basis of the description of the present application; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of each embodiment of the present application.

Claims

1.A method for time sensitive network traffic scheduling based on 5G network, characterized in that: The traffic scheduling method comprises: Data flow classification and priority sorting, detailed analysis and classification of incoming data flow, priority sorting according to the time sensitivity of the data flow; Configure time channel, allocate time channel with corresponding bandwidth and delay characteristics according to the category and priority of the data flow, ensure that the data flow with high time sensitivity has sufficient resources for transmission; Data flow scheduling, develop scheduling strategy, schedule data according to the time channel and priority of each type of data flow, ensure that the time requirements of each type of data flow are met; Configure node forwarding rules and load balancing strategy, configure forwarding rules for each node in the network, so that data can be transmitted according to the predetermined time channel and priority during transmission; Abnormal processing, when the network is abnormal, the system will process based on the preset processing strategy. 2.The method of claim 1, wherein: The data flow classification and priority sorting comprises: S101, determine the data flow to be classified; S102, define the characteristics of the data flow, including protocol type, service type, QoS requirement; S103, define classification rules, use decision tree model according to the characteristics and requirements of the data flow; S104, implement data flow classification, classify data flow according to its characteristics and defined classification rules; S105, priority sorting of classification results, define different priorities according to the results and the importance or sensitivity of each type of data; S106, allocate channels for each classification of data flow, allocate different time channels for classified data flow according to priority. 3.The method of claim 1, wherein: The configuration of time channel comprises: S201, creating time channels: for each type of data flow, a time channel j is created according to its priority, where n is the number of time channels; S202, allocate bandwidth: allocate bandwidth of each time channel based on priority and time sensitivity requirement of each data flow; bandwidth allocation strategy is represented as a function , wherein is the bandwidth allocated to time channel j; S203, setting delay characteristics of the channel: setting a lower network delay for a data stream with a higher priority; setting the delay characteristics is expressed as a function g, , wherein is a delay of the time channel j; S204, dynamically adjust the time channel attribute: dynamically adjust the bandwidth and delay of the time channel according to the real-time load state of the network, reduce the bandwidth of the time channel with low priority during traffic peak period to ensure that the key service is not affected. 4.The method of claim 1, wherein: The scheduling strategy comprises: S301, sort all data streams according to the arrival time a[i] of each data stream, and obtain a new sequence E', so that ; S302、At any time point t, calculate the execution time of all data streams that have arrived but have not started execution , and select the one with the shortest execution time to process, find the data stream i that satisfies ; S303, start processing the selected data streams, define a function for data stream i for calculating and updating the completion time of each data stream; S304、When all data streams are processed, that is The scheduling process ends when this is true for all i. 5.The method of claim 1, wherein: The configuration of node forwarding rules comprises: S401, calculate the response time of each node: send a small request to each node, and then estimate the actual response time according to the response time; S402, find the node with the shortest response time in all nodes: complete by comparing the response time of each node; preferentially select the node with the shortest response time for data packet transmission and processing; S403, send data flow to the node with the shortest response time: when there are multiple data flows, select according to their priority and the node with the shortest response time; S404, update the response time of the node: when a node has processed a request, the response time of the node needs to be recalculated, and the data flow allocation is performed again according to the new response time. 6.The method of claim 1, wherein: The load balancing strategy comprises: Assume that m is the total number of requests, n is the number of servers, and k is the sampling number, here k=n; then, for the ith server, the probability of being selected is; P=the probability that the random number generated from the 1st request to the mth request falls within the reservoir range = ∑(the probability that the random number from the 1st to the mth request falls within the reservoir range); The steps are as follows: (1) Initialize a sample pool with size N; each bit in the sample pool represents a server; (2) When a request from a client arrives, generate a random number i in the range of [0, current total number of requests]; suppose the current request is the jth request in the aggregate, then generate the random number i; (3) If the random number i is in the range of the sample pool, then let the i th server handle the request; (4) According to the reservoir sampling rule, the probability of each server being selected should be the same; therefore, after n requests end, it is expected that each server will handle an equal number of requests; that is, for the i th server, the probability of being selected is 1 / n; in this way, overall load balancing is achieved.

Images