Service flow QoS link availability analysis method and system

Abstract

The invention discloses a service flow QoS link availability analysis method and system. The method comprises the following steps: constructing a graph structure of a network; obtaining all shortest paths in the graph structure by using a shortest path priority algorithm, and calculating a service performance index by combining the network state parameter of each link and the number of the shortest paths contained in each link; sorting the links in combination with the service performance indexes of the links and the current bandwidth utilization rate; sequentially deleting the links according to the sequence, and stopping when the maximum connected sub-graph proportion of the residual graph structures is lower than a set connection threshold value or the average value of the sizes of all the shortest paths in the graph structures after the links are deleted is higher than a set path threshold value; and calculating the Rich curvature of the nodes at the two ends of each deleted link, so as to correct the service performance indexes of the links, and the corrected service performance indexes are final link availability. According to the method, multi-dimensional parameters are fused, and not only are the characteristics of the links considered, but also the topological characteristics of the links to the whole network are considered.

Description

Technical Field

[0001] The present invention belongs to the technical field of network traffic analysis, and more specifically, relates to a method and system for analyzing the availability of QoS links for service traffic. Background Art

[0002] With the rapid development of network technology and the increasing complexity of business requirements, traditional network quality of service (QoS) management methods are facing huge challenges. Existing QoS management technologies usually rely on single or limited network performance metrics, such as bandwidth, latency, data packet loss rate, etc., to evaluate and optimize network performance. Although this single-dimensional analysis method performs well in simple network environments, its effectiveness is limited in today's increasingly complex network environments. For example, the simple bandwidth utilization rate cannot comprehensively reflect the network congestion status, and the analysis of latency and data packet loss rate cannot accurately reflect the network stability. Therefore, a method that can comprehensively consider multiple performance metrics and comprehensively evaluate the network status is needed. Secondly, existing technologies have deficiencies in real-time performance and security. With the rapid change of service traffic, the network environment is also constantly changing, which requires the QoS management method to be able to respond to the changes in network status in real time, and the network should still be able to maintain its core functions when facing link failures, malicious attacks or burst traffic.

[0003] In the prior art, the methods for analyzing the availability of a network are as follows: CN105703973A, a QoS routing method based on a genetic algorithm, the steps of which are as follows: first, initialize all communication nodes, secondly, calculate the fitness of the entire sample population, then determine whether the maximum optimization algebra is reached. If the maximum evolution algebra is reached, the optimal solution can be confirmed. Otherwise, perform genetic operations to obtain the first link. At the same time, use the weighted breadth-first search method to obtain the second link, select between the two links, set the weight of each parameter according to the three parameters of the bandwidth, delay and packet loss rate of the link, and calculate the results of the weight calculation equations of the two communication links respectively, and select the link with the result value for network communication. This invention does not consider the redundancy and anti-interference ability of the link, nor does it analyze the link in combination with the entire network structure, and only sets according to the three parameters of the bandwidth, delay and packet loss rate of the link. Summary of the Invention

[0004] To solve the deficiencies existing in the prior art, the present invention provides a method and system for analyzing the availability of QoS links for service traffic.

[0005] The present invention adopts the following technical solutions.

[0006] In the first aspect of the present invention, a method for analyzing the availability of QoS links for service traffic is proposed, which is characterized by including: Collect network status parameters between each pair of links in the network, and construct a graph structure of the network; use the shortest path first algorithm to obtain all the shortest paths in the graph structure, and calculate the service performance metrics by combining the network status parameters of each link and the number of times each link is included in all the shortest paths; Sort the links according to the service performance metrics of the links and the current bandwidth utilization rate; Delete the links in sequence according to this sorting, calculate the proportion of the largest connected subgraph of the remaining graph structure, and the average value of the sizes of all the shortest paths in the graph structure after deleting this link; stop when the proportion of the largest connected subgraph of the remaining graph structure is lower than the set connectivity threshold or the average value of the sizes of all the shortest paths in the graph structure after deleting the link is higher than the set path threshold, and count all the deleted links; Calculate the discretized Ricci curvature of the two end nodes of each of these deleted links, and use the Ricci curvature to correct the service performance metrics of these links. The corrected service performance metrics are the final link availability.

[0007] Preferably, the collecting network status parameters between each pair of links in the network and constructing a graph structure of the network are specifically as follows: The network status parameters between each pair of links include bandwidth, delay, and packet loss rate; the graph structure is an undirected graph, the nodes of this graph structure are the transmission nodes of the network, and the edges are the links of the network.

[0008] Preferably, the using the shortest path first algorithm to obtain all the shortest paths in the graph structure is specifically as follows: According to the network structure and application scenario, set several source nodes and several target nodes; for any one source node; set the distance between the source node and itself to 0, and calculate the distance between the source node and all other nodes and set it to a maximum threshold, and the maximum threshold is at least one order of magnitude higher than the maximum distance of all nodes; Select the node with the smallest current distance among all the nodes, calculate the distances from this node to all its adjacent nodes, and thus calculate the distances from the source node to these adjacent nodes. If it is less than the previously set or calculated distances from the source node to these adjacent nodes, then update this distance to the currently calculated distance; repeat the above steps starting from the node with the smallest current distance among all the nodes until the node with the smallest current distance among all the nodes has no adjacent nodes; Obtain the distances from the source node to each target node at this time, where the shortest distance is the shortest path distance of this source node, and all the nodes passed by this distance are the shortest paths, and count the links included in all the shortest paths.

[0009] Preferably, the setting several source nodes and several target nodes according to the network structure and application scenario is specifically as follows: All edge nodes are set as source nodes. If the application scenario is data upload, the target node is set as the node connected to the server; if the application scenario is content distribution, the target node is set as the node connected to the user terminal; if the application scenario is collaborative computing, the target node is also an edge node. For a certain edge node as the source node, the target node must be other edge nodes and cannot be the same as the source node; if the application scenario is dynamic optimization of resource allocation, then select the node with the largest bandwidth or the lowest latency as the target node according to the actual business requirements.

[0010] Preferably, the service performance index is calculated by combining the network state parameters of each link and the number of times each link is included in all the shortest paths. , the formula is:

[0011] Wherein, represents the bandwidth; represents the set maximum bandwidth; represents the latency; represents the reference latency; represents the packet loss rate; represents the packet loss rate threshold; represents the number of times each link is included in all the shortest paths; represents the smoothing coefficient.

[0012] Preferably, the links are sorted by combining the service performance index of the link and the current bandwidth utilization rate. Specifically: Normalize the service performance index of the link and subtract the bandwidth utilization rate from 1; If in the normalized service performance index of link A and 1 minus the bandwidth utilization rate, any one is higher than link B by a set first threshold, and the other is not lower than link B by a set second threshold, where the first threshold is twice the second threshold, then it is considered that link A dominates link B; analyze the dominance relationship between every two links, and sort the links from high to low according to the dominance number; for links with the same dominance number, they are sorted from large to small according to the historical failure rate of the links.

[0013] Preferably, the proportion of the largest connected subgraph of the remaining graph structure is the proportion of the total number of links included in the largest connected subgraph in the remaining graph structure to the total number of links in the original graph structure; the connection threshold is set to 40%; The path threshold is set to twice the average value of the sizes of all the shortest paths in the original graph structure.

[0014] Preferably, the discretized Ricci curvature of each end node of each of these deleted links is calculated. Specifically: Calculate the local probability distribution of the two end nodes of the link. For the node Local probability distribution is as follows:

[0015] Wherein, is the set of all adjacent nodes of represents the union of and is the number of all adjacent nodes of z is any node in the graph structure and is the independent variable of the local probability distribution; Calculate the optimal transport distance 1-Wasserstein distance between two local probability distributions; the discretized Ricci curvature of the nodes at both ends of the link is equal to 1 minus the 1-Wasserstein distance.

[0016] Preferably, the service performance index of these links is corrected using the Ricci curvature, specifically: The formula for the corrected service performance index is:

[0017] Wherein, is the corrected service performance index; is the original service performance index; is the Ricci curvature; , are both set correction coefficients.

[0018] The second aspect of the present invention proposes a business traffic QoS link availability analysis system using the method described in the first aspect of the present invention, including a graph structure construction module, a service performance index calculation module, a sorting module, an important link screening module, and a service performance index correction module, characterized in that: Graph structure construction module: Collect network state parameters between each link in the network and construct the graph structure of the network; Service performance index calculation module: Use the shortest path first algorithm to obtain all shortest paths in the graph structure, and calculate the service performance index in combination with the network state parameters of each link and the number of times each link is included in all shortest paths; Sorting module: Sort the links in combination with the service performance index of the links and the current bandwidth utilization rate; Important link screening module: Delete the links in sequence according to this sorting, calculate the proportion of the largest connected subgraph of the remaining graph structure, and the average value of the sizes of all the shortest paths in the graph structure after deleting this link; stop when the proportion of the largest connected subgraph of the remaining graph structure is lower than the set connectivity threshold or the average value of the sizes of all the shortest paths in the graph structure after deleting the link is higher than the set path threshold, and count all the deleted links; Service performance index correction module: Calculate the discretized Ricci curvature of the two end nodes of each link among all these deleted links, and use the Ricci curvature to correct the service performance index of these links. The corrected service performance index is the final link availability.

[0019] The beneficial effects of the present invention are as follows. Compared with the prior art, the service performance index is calculated by combining the network state parameters of each link and the number of times each link is included in all the shortest paths, integrating multi-dimensional parameters, and considering not only the characteristics of the link but also the topological characteristics of the link to the entire network; the links are sorted by combining the service performance index of the link, the current bandwidth utilization rate, and the failure rate, and the dynamic real-time characteristics of the link are considered by using the current bandwidth utilization rate; the links are deleted in sequence according to this sorting, and stop when the threshold is reached, and count all the deleted links; these links are the key links that are important to the entire network, have high utilization rate and are more likely to be utilized, and are prone to failure. Deleting these will cause the network to collapse; only the indexes of these links are corrected; avoiding introducing the curvature of non-critical links to add noise and causing overcorrection and increasing the calculation amount; correcting the critical links with discretized Ricci curvature to capture the vulnerability or redundancy of the link in the topological structure. Description of the Drawings

[0020] Figure 1 It is a flowchart of the present invention. Detailed Embodiments

[0021] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. The embodiments described in this application are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the spirit of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present invention.

[0022] As Figure 1 shown, Embodiment 1 of the present invention proposes a method for analyzing the link availability of service traffic QoS, which is characterized by including: Collect network status parameters between each pair of links in the network and construct the graph structure of the network; use the shortest path first algorithm to obtain all the shortest paths in the graph structure, and calculate the service performance metrics by combining the network status parameters of each link and the number of times each link is included in all the shortest paths; Sort the links by combining the service performance metrics of the links and the current bandwidth utilization rate; Delete the links in sequence according to this sorting, calculate the proportion of the largest connected subgraph of the remaining graph structure, and the average value of the sizes of all the shortest paths in the graph structure after deleting this link; stop when the proportion of the largest connected subgraph of the remaining graph structure is lower than the set connectivity threshold or the average value of the sizes of all the shortest paths in the graph structure after deleting the link is higher than the set path threshold, and count all the deleted links; Calculate the discretized Ricci curvature of the two end nodes of each of these deleted links, and use the Ricci curvature to correct the service performance metrics of these links. The corrected service performance metrics are the final link availability.

[0023] The collecting of the network status parameters between each pair of links in the network and the construction of the graph structure of the network are specifically as follows: The network status parameters between each pair of links include bandwidth, delay, and packet loss rate; the graph structure is an undirected graph, the nodes of this graph structure are the transmission nodes of the network, and the edges are the links of the network.

[0024] The using of the shortest path first algorithm to obtain all the shortest paths in the graph structure is specifically as follows: According to the network structure and application scenario, set a number of source nodes and a number of target nodes; for any one source node; set the distance between the source node and itself to 0, and calculate the distance between the source node and all other nodes and set it to a maximum threshold, and the maximum threshold is at least one order of magnitude higher than the maximum distance of all nodes; Take out the node with the smallest current distance among all the nodes, calculate the distances from this node to all its adjacent nodes, so as to calculate the distances from the source node to these adjacent nodes. If it is less than the previously set or calculated distances from the source node to these adjacent nodes, then update this distance to the currently calculated distance; repeat the above steps starting from taking out the node with the smallest current distance among all the nodes until the node with the smallest current distance among all the nodes has no adjacent nodes; Obtain the distances from the source node to each target node at this time, where the shortest distance is the shortest path distance of this source node, and all the nodes passed by this distance are the shortest paths, and count the links included in all the shortest paths.

[0025] The setting of a number of source nodes and a number of target nodes according to the network structure and application scenario is specifically as follows: All edge nodes are set as source nodes. If the application scenario is data upload, the destination node is set as the node connected to the server; if the application scenario is content distribution, the destination node is set as the node connected to the user terminal; if the application scenario is collaborative computing, the destination node is also an edge node. For an edge node as a source node, the destination node must be other edge nodes and cannot be the same as the source node; if the application scenario is dynamic optimization of resource allocation, then select the node with the largest bandwidth or the lowest latency as the destination node according to the actual business requirements.

[0026] Calculate the service performance index by combining the network state parameters of each link and the number of times each link is included in all the shortest paths. , the formula is:

[0027] Among them, represents the bandwidth; represents the set maximum bandwidth; represents the latency; represents the reference latency; represents the packet loss rate; represents the packet loss rate threshold; represents the number of times each link is included in all the shortest paths; represents the smoothing coefficient.

[0028] Specifically, the smoothing coefficient of this embodiment is set to 1.5.

[0029] Sort the links by combining the service performance index of the link and the current bandwidth utilization rate. Specifically: Normalize the service performance index of the link and subtract the bandwidth utilization rate from 1; If in the normalized service performance index of link A and 1 minus the bandwidth utilization rate, either one is higher than link B by a set first threshold, and the other is not lower than link B by a set second threshold, where the first threshold is twice the second threshold, then it is considered that link A dominates link B; analyze the dominance relationship between every two links and sort the links in descending order of the dominance number; for links with the same dominance number, sort them in descending order of the historical failure rate of the link.

[0030] The proportion of the largest connected subgraph of the remaining graph structure is the proportion of the total number of links included in the largest connected subgraph in the remaining graph structure to the total number of links in the original graph structure; the connectivity threshold is set to 40%; The path threshold is set to twice the average size of all the shortest paths in the original graph structure.

[0031] Calculate the discretized Ricci curvature of the two end nodes of each link among all these deleted links. Specifically: Calculate the local probability distributions of the nodes at both ends of the computing link. For a node the local probability distribution is:

[0032] where is the set of all adjacent nodes of denotes the union of and is the number of all adjacent nodes of z is any node in the graph structure and is the independent variable of the local probability distribution; Calculate the optimal transport distance, the 1-Wasserstein distance, between the two local probability distributions; the discretized Ricci curvature of the nodes at both ends of the link is equal to 1 minus the 1-Wasserstein distance.

[0033] Specifically, the 1-Wasserstein distance is calculated as:

[0034] where represents the transport plan for transporting node u to node v i.e., the proportion of resources flowing from node u to node v ; is the set of all adjacent nodes of node denotes the union of node and denotes the with the minimum transport cost; denotes u the shortest path from node v to node denotes the local probability distribution of node , are the probabilities obtained by substituting node u , node v into the local probability distributions of nodes , respectively.

[0035] The formula for the discretized Ricci curvature of the nodes at both ends of the link is:

[0036] Using the Ricci curvature to correct the service performance metrics of these links, specifically: The formula for the corrected service performance metric is:

[0037] Where, is the corrected service performance metric; is the original service performance metric; is the Ricci curvature; , are both set correction coefficients.

[0038] Specifically, is set to 0.5, is set to 2.0.

[0039] Embodiment 2 of the present invention proposes a business traffic QoS link availability analysis system using the method described in Embodiment 1 of the present invention, including a graph structure construction module, a service performance metric calculation module, a sorting module, an important link screening module, and a service performance metric correction module, characterized in that: Graph structure construction module: Collect network state parameters between each link in the network and construct a graph structure of the network; Service performance metric calculation module: Use the shortest path first algorithm to obtain all shortest paths in the graph structure, and calculate the service performance metric by combining the network state parameters of each link and the number of times each link is included in all shortest paths; Sorting module: Sort the links by combining the service performance metric of the link and the current bandwidth utilization rate; Important link screening module: Delete links in sequence according to this sorting, calculate the proportion of the largest connected subgraph of the remaining graph structure, and the average value of the sizes of all shortest paths in the graph structure after deleting the link; Stop when the proportion of the largest connected subgraph of the remaining graph structure is lower than the set connectivity threshold or the average value of the sizes of all shortest paths in the graph structure after deleting the link is higher than the set path threshold, and count all the deleted links; Service performance metric correction module: Calculate the discretized Ricci curvature of the two end nodes of each of these deleted links, use the Ricci curvature to correct the service performance metrics of these links, and the corrected service performance metric is the final link availability.

[0040] The present disclosure may be a system, method, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions thereon for causing a processor to implement various aspects of the present disclosure.

[0041] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: it is still possible to modify the specific implementation manners of the present invention or make equivalent replacements, and any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A method for analyzing the availability of a service traffic QoS link, characterized in that, Including: Collecting network state parameters between each pair of links in the network and constructing a graph structure of the network; Using the shortest path first algorithm to obtain all the shortest paths in the graph structure, and calculating service performance metrics by combining the network state parameters of each link and the number of times each link is included in all the shortest paths; Sorting the links by combining the service performance metrics of the links and the current bandwidth utilization rate; Deleting the links in sequence according to this sorting, calculating the proportion of the largest connected subgraph of the remaining graph structure, and the average value of the sizes of all the shortest paths in the graph structure after deleting this link; stopping when the proportion of the largest connected subgraph of the remaining graph structure is lower than the set connectivity threshold or the average value of the sizes of all the shortest paths in the graph structure after deleting the link is higher than the set path threshold, and counting all the deleted links; Calculating the discretized Ricci curvature of the two end nodes of each of these deleted links, and using the Ricci curvature to correct the service performance metrics of these links, and the corrected service performance metrics are the final link availability.

2. A method for analyzing the link availability of service traffic QoS according to claim 1, characterized in that: The step of collecting network state parameters between each pair of links in the network and constructing a graph structure of the network is specifically: The network state parameters between each pair of links include bandwidth, delay, and packet loss rate; the graph structure is an undirected graph, the nodes of this graph structure are the transmission nodes of the network, and the edges are the links of the network.

3. A method for analyzing the link availability of service traffic QoS according to claim 2, characterized in that: The step of using the shortest path first algorithm to obtain all the shortest paths in the graph structure is specifically: According to the network structure and application scenario, setting a number of source nodes and a number of target nodes; for any one source node, setting the distance between the source node and itself to 0, and calculating the distance between the source node and all other nodes to be set as a maximum threshold, and the maximum threshold is at least one order of magnitude higher than the maximum distance of all nodes; Taking out the node with the smallest current distance among all the nodes, calculating the distances from this node to all its adjacent nodes, thereby calculating the distances from the source node to these adjacent nodes, and if it is less than the previously set or calculated distances from the source node to these adjacent nodes, then updating this distance to the currently calculated distance; repeating the above steps starting from taking out the node with the smallest current distance among all the nodes until there are no adjacent nodes for the node with the smallest current distance taken out among all the nodes; Obtaining the distances from the source node to each target node at this time, where the shortest distance is the shortest path distance of this source node, and all the nodes passed through corresponding to this distance are the shortest paths, and counting the links included in all the shortest paths.

4. A method for analyzing the link availability of service traffic QoS according to claim 3, characterized in that: The step of setting a number of source nodes and a number of target nodes according to the network structure and application scenario is specifically: All edge nodes are set as source nodes. If the application scenario is data upload, the target node is set as the node connected to the server; if the application scenario is content distribution, the target node is set as the node connected to the user terminal; if the application scenario is collaborative computing, the target node is also an edge node. When a certain edge node is used as the source node, the target node must be other edge nodes and cannot be the same as the source node; if the application scenario is dynamic optimization of resource allocation, then the node with the largest bandwidth or the lowest latency is selected as the target node according to the actual business requirements.

5. A method for analyzing the availability of a service traffic QoS link according to claim 3, characterized in that: Calculate the service performance metrics by combining the network status parameters of each link and the number of times each link is included in all the shortest paths , and the formula is as follows: Among them, represents the bandwidth; represents the set maximum bandwidth; represents the delay; represents the reference delay; represents the packet loss rate; represents the packet loss rate threshold; represents the number of times each link is included in all the shortest paths; represents the smoothing coefficient.

6. A method for analyzing the availability of a service traffic QoS link according to claim 3, characterized in that: Sorting the links by combining the service performance indicators of the links and the current bandwidth utilization rate, specifically: Normalize the service performance indicators of the links and subtract the bandwidth utilization rate from 1; If, in the normalized service performance indicators of link A and 1 minus the bandwidth utilization rate, any one is higher than link B by a set first threshold, and the other is not lower than link B by a set second threshold, where the first threshold is twice the second threshold, then it is considered that link A dominates link B; analyze the dominance relationship between every two links, and sort the links in descending order of the dominance number; for links with the same dominance number, they are sorted in descending order of the historical failure rate of the links.

7. A method for analyzing the availability of a service traffic QoS link according to claim 1, characterized in that: The proportion of the largest connected subgraph of the remaining graph structure is the proportion of the total number of links included in the largest connected subgraph in the remaining graph structure to the total number of links in the original graph structure; the connection threshold is set to 40%; The path threshold is set to twice the average of the sizes of all the shortest paths of the original graph structure.

8. A method for analyzing the availability of a service traffic QoS link according to claim 3, characterized in that: Calculating the discretized Ricci curvature of the two end nodes of each of these deleted links, specifically: Calculate the local probability distributions of the nodes at both ends of the computing link. For node the local probability distribution is as follows: Among them, is the set of all adjacent nodes of; denotes the union of and; is the number of all adjacent nodes of, z is any node in the graph structure and is the independent variable of the local probability distribution; Calculating the optimal transport distance, the 1-Wasserstein distance, between two local probability distributions; the discretized Ricci curvature of the two end nodes of the link is equal to 1 minus the 1-Wasserstein distance.

9. A method for analyzing the availability of a service traffic QoS link according to claim 1, characterized in that: Using the Ricci curvature to correct the service performance indicators of these links, specifically: The formula for the corrected service performance indicator is: Among them, is the corrected service performance index; is the original service performance index; is the Ricci curvature; and are both set correction coefficients.

10. A service traffic QoS link availability analysis system using the method according to any one of claims 1-9, including a graph structure construction module, a service performance indicator calculation module, a sorting module, a critical link screening module, and a service performance indicator correction module, characterized in that: The graph structure construction module: collects the network state parameters between each link in the network and constructs the graph structure of the network; Service performance metric calculation module: Obtain all the shortest paths in the graph structure using the shortest path first algorithm, and calculate the service performance metrics by combining the network status parameters of each link and the number of times each link is included in all the shortest paths; Sorting module: Sort the links by combining the service performance metrics of the links and the current bandwidth utilization rate; Important link screening module: Delete the links in sequence according to this sorting, calculate the proportion of the largest connected subgraph of the remaining graph structure, and the average value of the sizes of all the shortest paths in the graph structure after deleting this link; Stop when the proportion of the largest connected subgraph of the remaining graph structure is lower than the set connectivity threshold or the average value of the sizes of all the shortest paths in the graph structure after deleting the link is higher than the set path threshold, and count all the deleted links; Service performance metric correction module: Calculate the discretized Ricci curvature of the two end nodes of each of these deleted links, and use the Ricci curvature to correct the service performance metrics of these links. The corrected service performance metric is the final link availability.

Images