Wide area network flow scheduling method for sensing link congestion

Abstract

According to the wide area network flow scheduling method for sensing link congestion disclosed by the invention, the congestion hotspot link in the network is identified in advance by analyzing the link congestion condition, the virtual link capacity is generated by scaling the link capacity of the congestion hotspot link, and the flow scheduling for coping with the network fluctuation is carried out based on the virtual link capacity. According to the method, the traffic is enabled to avoid the congestion hotspot link, so that a buffer space is provided for the traffic fluctuation, the robustness to the traffic fluctuation is improved, a subsequent routing strategy is optimized by generating the virtual link capacity, and a specific traffic engineering scheme does not need to be bound, so that the method can be seamlessly integrated into various existing traffic engineering schemes; only a small number of congestion hotspot links are intervened, so that the utilization rate, namely the median performance, under the conventional condition is slightly influenced while the performance under the worst condition is greatly improved to deal with burst traffic.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of computer networks, and particularly relates to a wide area network traffic scheduling method for perceiving link congestion. BACKGROUND

[0002] With the rise of cloud computing and big data services, the traffic in the wide area network presents highly dynamic and high burst characteristics. Traffic engineering technology aims to maximize network resource utilization and reduce congestion by optimizing routing paths and traffic distribution.

[0003] The existing traffic engineering scheme for obtaining routing strategies generally includes two steps: first, model the routing problem as a linear programming problem, such as a multi-commodity flow model; and then use a linear programming solver to solve the linear programming problem to generate a routing strategy for each traffic matrix. However, the routing strategies generated by the algorithms proposed in existing research are generally for fixed traffic matrices. When network fluctuations occur, the traffic matrix changes greatly, and the routing strategy generated for the historical traffic matrix no longer fits the current traffic matrix, thereby causing a decline in network performance.

[0004] In actual large-scale wide area network operation, network congestion often does not occur uniformly on all links, but is concentrated on a small number of links, and the congestion conditions between links differ greatly. Some links located in the network core or critical path are prone to become bottlenecks due to burst traffic, while other links may be in a light load state for a long time. When the network encounters unforeseen traffic fluctuations, these already near-saturated congestion hotspot links lack sufficient elastic margin to buffer traffic impact, and are extremely prone to cause serious congestion, seriously affecting network performance. Therefore, if the congestion hotspot links that are prone to become bottlenecks can be accurately identified, and preventive traffic avoidance measures are taken, it will be of great significance to improve the overall robustness of the network.

[0005] Therefore, how to effectively reserve the resources of the congestion hotspot link under the premise of ensuring computational efficiency and prevent it from becoming a bottleneck for load balancing is a problem to be solved in the current traffic engineering field. SUMMARY

[0006] Therefore, the application provides a wide area network traffic scheduling method for perceiving link congestion, which solves the problem of decline in network anti-fluctuation ability caused by premature saturation of congestion hotspot links in the existing routing strategy.

[0007] The wide area network traffic scheduling method for perceiving link congestion provided by the application specifically includes the following steps:

[0008] Step 1, based on a set of historical traffic matrices, calculate the characteristic indexes of each link in the network, and identify the congestion hotspot link through a preset screening rule;

[0009] Step 2, multiplying the link capacity of the identified congestion hotspot link by a scaling factor less than 1 to obtain a scaled virtual link capacity; for non-congestion hotspot links, the virtual link capacity thereof remains unchanged;

[0010] Step 3, generating a routing strategy based on a path-based multi-commodity flow model based on the virtual link capacity.

[0011] Further, the preset screening rules are a high link utilization rate rule and a high link endpoint degree rule.

[0012] Further, the specific calculation method of the high link utilization rate rule is as follows:

[0013] Calculate the average traffic matrix of the historical traffic matrix set, assume that the traffic is uniformly distributed on the paths of all source and destination pairs, and calculate the average utilization rate Lu of each link i ; if the utilization rate Lu of link k k satisfies the following condition, it is determined to be a potential congestion hotspot link:

[0014]

[0015] Lu i >α·Lu avg

[0016] wherein T i is the average traffic allocated to link i under the condition that the historical traffic matrix is uniformly distributed on the paths of all source and destination pairs, C i is the load capacity of link i, K is the total number of links in the network, and a is a preset utilization rate threshold coefficient, a is an adjustable parameter greater than 1.

[0017] Further, the specific calculation method of the high link endpoint degree rule is as follows:

[0018] For a link satisfying the high link utilization rate rule, check the number of links connected to its two end nodes as the degree, if the sum of the degrees of the source node and the destination node of link k satisfies the following condition, it is confirmed to be a congestion hotspot link:

[0019]

[0020] wherein is the degree of the source node , is the degree of the destination node .

[0021] Further, the calculation process of the virtual link capacity is as follows:

[0022]

[0023] wherein C i is the load capacity of link i, and γ is the link scaling factor.

[0024] Further, the path-based multi-commodity flow model is:

[0025] min PR

[0026]

[0027] e∈E c ,D s,d ∈TM

[0028] wherein PR is the performance gap between the current routing policy and the optimal routing policy, is the proportion of traffic demand from source s to destination d allocated to path p, and P s,d is the set of paths between source s and destination d, V is the set of network nodes, Bk e is the traffic of non-critical flow on link e, is the path partition rate, D s,d is the traffic demand from source s to destination d, c e is the capacity of link e, U opt (TM) is the maximum link utilization rate that can be achieved under the traffic matrix condition, E c is the set of selected critical links.

[0029] Further, the path-based multi-commodity flow model is solved by using an LP optimizer to obtain a routing policy.

[0030] Further, the path-based multi-commodity flow model is solved by using a mathematical programming solver Gurobi to obtain a routing policy, specifically:

[0031]

[0032] wherein σ t-1 represents the routing policy generated at time t-1, TM t-1 represents the traffic matrix collected at time t-1, and MCF(·) represents the solving process of the path-based multi-commodity flow model.

[0033] Beneficial effects are:

[0034] The application pre-identifies the congestion hotspot link in the network by analyzing the link congestion situation, generates virtual link capacity by scaling the link capacity of the congestion hotspot link, and schedules the traffic based on the virtual link capacity to cope with the traffic fluctuation, so as to provide a buffer space for the traffic fluctuation and improve the robustness to the traffic fluctuation. Since the virtual link capacity is generated to optimize the subsequent routing strategy, it is not necessary to bind a specific traffic engineering scheme, so it can be seamlessly integrated into various existing traffic engineering schemes. In addition, by only intervening in a small number of congestion hotspot links, the utilization rate under normal circumstances, i.e. the median performance, is less affected while the worst-case performance is greatly improved to cope with burst traffic. BRIEF DESCRIPTION OF DRAWINGS

[0035] Figure 1 A processing flow diagram of a wide area network traffic scheduling method for sensing link congestion provided by the application.

[0036] Figure 2 A schematic diagram of congestion hotspot link identification logic in a wide area network traffic scheduling method for sensing link congestion provided by the application. DETAILED DESCRIPTION

[0037] The application will be described in detail below with specific embodiments.

[0038] The wide area network traffic scheduling method for sensing link congestion provided by the application has the core idea that: by analyzing the link congestion situation, virtual link capacity is generated, and traffic scheduling is performed based on the adjusted virtual link capacity. First, the topology congestion hotspot link is judged by HLU and HDE criteria, then the link capacity of the congestion hotspot link is scaled to generate virtual link capacity, and the routing strategy is generated by the traffic engineering scheme based on the capacity. On the one hand, the virtual link capacity generated by the scheme is independent of a specific algorithm and can be used with other traffic engineering schemes. On the other hand, the scheme greatly improves the worst-case performance while having less impact on the median performance.

[0039] The wide area network traffic scheduling method for sensing link congestion provided by the application has the core idea that: by analyzing the link congestion situation, virtual link capacity is generated, and traffic scheduling is performed based on the adjusted virtual link capacity. First, the topology congestion hotspot link is judged by HLU and HDE criteria, then the link capacity of the congestion hotspot link is scaled to generate virtual link capacity, and the routing strategy is generated by the traffic engineering scheme based on the capacity. On the one hand, the virtual link capacity generated by the scheme is independent of a specific algorithm and can be used with other traffic engineering schemes. On the other hand, the scheme greatly improves the worst-case performance while having less impact on the median performance. Figure 1 As shown in the figure, the method specifically includes the following steps:

[0040] Step 1, obtaining network topology and historical traffic data, identifying congestion hotspot links, as shown in Figure 2

[0041] Based on the historical traffic matrix set, the characteristic indexes of each link in the network are calculated, and the congestion hotspot links that are easy to become load balancing bottlenecks are identified through a pre-set screening rule.

[0042] ​Among them, congested hotspot links refer to those links that are highly susceptible to becoming load balancing bottlenecks during traffic fluctuations. This invention uses High Link Utilization (HLU) rules and High Link Endpoint Degree (HDE) rules to locate congested hotspot links:

[0043] (1) High Link Utilization (HLU) rule.

[0044] Calculate the average flow matrix TM based on the historical flow matrix in the training set. avg Assuming traffic is evenly distributed across all pre-configured paths, calculate the utilization Lu for each link i. i Then calculate the average utilization of all links. Set a threshold coefficient α, if link i satisfies Lu i >α·Lu avg If so, the link is marked as a potential congestion hotspot link.

[0045] (2) High Link Endpoint Degree (HDE) rule.

[0046] For links that meet the high link utilization rule, their topological location should be further examined. Links connecting high-density nodes typically carry paths and are more susceptible to fluctuations.

[0047] Obtain the degree of the source node S and destination node D of link i, i.e., the number of connected links;

[0048] If a degree threshold β is set, and degree(S) + degree(D) > β, then the link is ultimately identified as a congested hotspot link.

[0049] Step 2: Scaling up the capacity of congested hotspot links to generate virtual link capacity.

[0050] Define the link capacity vector C = C1, C2, ..., C K For each identified congested hotspot link, before performing the first-stage route calculation, its capacity is scaled to generate virtual link capacity. The scaling method is as follows:

[0051]

[0052] Here, γ is a scaling factor, with a value less than 1. For normal links, the capacity remains unchanged.

[0053] Step 3: Generate routing policies based on virtual link capacity.

[0054] Virtual link capacity As a subsequent traffic engineering solution, such as the link capacity of a path-based multi-commodity flow model, the historical traffic matrix is ​​input, a routing strategy is generated, and a solver is used to solve it.

[0055] Specifically, the path-based multi-product flow model is as follows:

[0056] min PR

[0057]

[0058] e∈E c D s,d ∈TM

[0059] Where PR represents the performance difference between the current routing policy and the optimal routing policy. P represents the proportion of traffic demand from source s to destination d that is allocated to path p. s,d Let V be the set of paths from source s to target d, and Bk be the set of network nodes. e For non-critical flows on link e, D represents the path segmentation rate. s,d For the traffic demand from source s to destination d, c e For the capacity of link e, U opt (TM) represents the maximum link utilization achievable under the traffic matrix condition, E c This represents the selected set of critical links. The routing strategy can be obtained by solving this model using the LP optimizer.

[0060] Furthermore, the present invention can also use the mathematical programming solver Gurobi to obtain the optimal routing strategy:

[0061]

[0062] Where, σ t-1 TM represents the routing policy generated at time t-1. t-1 represents the flow matrix collected at time t-1, and MCF(·) represents the solution process of the path-based multi-commodity flow model.

[0063] The performance of the proposed method was evaluated through experimental simulation. Compared with the original traffic engineering scheme that did not use the link scaling strategy, the present invention can improve the worst performance by 2.1 times while maintaining a similar performance level.

[0064] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0065] Finally, it should be noted that those skilled in the art can obviously make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims and their equivalents, this invention also intends to include these modifications and variations.

[0066] The above description is merely one embodiment of the present invention, and should not be construed as limiting the scope of the invention. Any structural changes made based on the present invention, as long as they do not depart from the essence of the invention, should be considered as falling within the protection scope of the present invention and subject to its restrictions. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process and related descriptions of the method described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0067] The term "comprising" or any other similar term is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus / method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent in such process, method, article, or apparatus / method.

[0068] The technical solutions of the present invention have been described above with further embodiments listed. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0069] In summary, the above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for wide area network traffic scheduling that senses link congestion, characterized in that, Specifically, the following steps are included: Step 1: Based on the historical traffic matrix set, calculate the characteristic indicators of each link in the network, and identify the congested hotspot links through preset filtering rules; Step 2: Multiply the link capacity of the identified congested hotspot links by a scaling factor less than 1 to obtain the scaled virtual link capacity. For non-congested hotspot links, their virtual link capacity remains unchanged; Step 3: Generate routing strategies using a path-based multi-commodity flow model based on virtual link capacity.

2. The wide area network traffic scheduling method according to claim 1, characterized in that, The preset filtering rules are the high link utilization rule and the high link endpoint degree rule.

3. The wide area network traffic scheduling method according to claim 2, characterized in that, The specific calculation method for the high link utilization rule is as follows: Calculate the average traffic matrix of the historical traffic matrix set, assuming that traffic is uniformly distributed across all source and destination pairs, and calculate the average utilization Lu for each link. i If the utilization rate of link k is Lu k A link is identified as a potential congestion hotspot if it meets the following conditions: Lu i >α·Lu avg Among them, T i C represents the average traffic distributed across link i under the condition that the historical traffic matrix is ​​uniformly distributed across all source and destination pairs. i Let be the load capacity of link i, K be the total number of links in the network, α be the preset utilization threshold coefficient, and α be an adjustable parameter greater than 1.

4. The wide area network traffic scheduling method according to claim 3, characterized in that, The specific calculation method for the high link endpoint degree rule is as follows: For links that meet the high link utilization rule, the number of links connected to both ends of the link is used as the degree. If the source node of link k... and destination node If the sum of the degrees of a link satisfies the following condition, it is confirmed as a congested hotspot link: in, For the source node The degree, destination node The degree.

5. The wide area network traffic scheduling method according to claim 1, characterized in that, The calculation process for the virtual link capacity is as follows: Among them, C i Let γ be the load capacity of link i, and γ be the link scaling factor.

6. The wide area network traffic scheduling method according to claim 1, characterized in that, The path-based multi-commodity flow model is as follows: min PR e∈E c ,D s,d ∈TM Where PR represents the performance difference between the current routing policy and the optimal routing policy. P represents the proportion of traffic demand from source s to destination d that is allocated to path p. s,d Let V be the set of paths from source s to target d, and Bk be the set of network nodes. e For non-critical flows on link e, D represents the path segmentation rate. s,d For the traffic demand from source s to destination d, c e For the capacity of link e, U opt (TM) represents the maximum link utilization achievable under the traffic matrix condition, E c The selected set of critical links.

7. The wide area network traffic scheduling method according to claim 6, characterized in that, The routing strategy is obtained by solving the path-based multi-commodity flow model using the LP optimizer.

8. The wide area network traffic scheduling method according to claim 6, characterized in that, The routing strategy is obtained by solving the path-based multi-commodity flow model using the mathematical programming solver Gurobi, specifically as follows: Where, σ t-1 TM represents the routing policy generated at time t-1. t-1 represents the flow matrix collected at time t-1, and MCF(·) represents the solution process of the path-based multi-commodity flow model.

Images