A method and system for assisting decision-making of cloud virtual network node deployment

Abstract

This invention provides an auxiliary decision-making method and system for deploying cloud virtual network nodes. The method includes: acquiring the forwarding latency data of each node in the entire set of candidate network nodes in the cloud; cleaning all latency data of each node; and weighted fusion of the cleaned data under different time periods and different network traffic to obtain the final latency data of each node matching the actual scenario; based on the final latency data of each node, obtaining the minimum connected graph under a preset network traffic; traversing the shortest path between any two nodes in the minimum connected graph based on the final latency data of each node; and using an improved variable neighborhood search algorithm to expand and search the neighborhood of the shortest path to obtain a network topology graph that satisfies the optimal latency and the degree of each node in the network topology graph. The above method optimizes the potential impact of traditional coarse-grained deployment on virtual network latency.

Description

Technical Field

[0001] This invention relates to the field of cloud computing technology, and in particular to an auxiliary decision-making method for deploying cloud virtual network nodes, an auxiliary decision-making system for deploying cloud virtual network nodes, an electronic device, and a computer-readable medium. Background Technology

[0002] With the rapid development of cloud computing, customers have increasingly higher requirements for network latency. The basic idea of ​​network virtualization in cloud computing is to explore hardware universality. Network virtualization optimization for transmission rate can be broadly divided into scheduling optimization and transmission optimization. Transmission optimization can be further divided into hardware device optimization and transmission line optimization. Transmission line optimization generally equates the bottleneck to the upper limit of the physical network, while often ignoring the differences in transmission rate between hundreds or thousands of physical devices.

[0003] Current deployment methods for cloud virtual networks typically involve mechanically planning the types and quantities of virtual network devices within a given set of physical devices, assuming that the transmission rates between these devices are completely identical, even though actual speeds vary. Deploying in this traditional way may present the following two problems:

[0004] (1) The transmission rate between two high-volume network elements is not high, which leads to the network's latency bottleneck due to deployment issues.

[0005] (2) Some high-volume network elements are not located in most of the shorter latency paths or are deployed on edge nodes. When they collect and distribute virtual machine traffic, they cannot provide the shortest average latency. Summary of the Invention

[0006] The present invention provides an auxiliary decision-making method for deploying cloud virtual network nodes, an auxiliary decision-making system for deploying cloud virtual network nodes, an electronic device, and a computer-readable medium to solve the problem in the prior art that network latency bottlenecks are limited by deployment issues and cannot achieve the shortest latency.

[0007] This invention discloses an auxiliary decision-making method for deploying cloud virtual network nodes, comprising:

[0008] Obtain the forwarding latency data of each node in the full set of candidate nodes in the cloud network. The latency data is the forwarding data of each node under each network traffic when the cloud virtual network goes online with the help of the streaming system.

[0009] All latency data for each node is cleaned, and the cleaned data from different time periods and under different network traffic conditions are weighted and fused to obtain the final latency data for each node that matches the actual scenario.

[0010] Based on the final latency data of each node, obtain the minimum connectivity graph under the preset network traffic. The minimum connectivity graph is a star topology structure that contains latency data within a specified range between any two nodes.

[0011] Based on the final latency data of each node, the shortest path between any two nodes in the minimum connected graph is traversed. An improved variable neighborhood search algorithm is used to expand and search the neighborhood of the shortest path to obtain a network topology graph that satisfies the optimal latency and the degree of each node in the network topology graph.

[0012] Optionally, obtain the forwarding latency data for each node in the entire set of candidate nodes in the cloud network, including:

[0013] By using a streaming system to simulate a cloud virtual network without load, the first latency data between any two nodes at different time intervals is obtained;

[0014] When the virtual network traffic of the streaming system reaches n specified network traffic values, the second latency data between any two nodes at different time intervals is obtained; n is greater than or equal to 1.

[0015] The first delay data and the second delay data constitute the delay data forwarded by each node.

[0016] Optionally, all latency data for each node is cleaned, and the cleaned data from different time periods and under different network traffic conditions are weighted and fused, including:

[0017] For all latency data of each node, a density-based outlier detection method is used to remove outliers from the latency data, and a k-nearest neighbor interpolation method is used to process noise in the latency data to obtain preprocessed latency data.

[0018] Based on the actual network traffic after the cloud virtual network is deployed, the preprocessed latency data under different times and network traffic conditions are weighted and merged to obtain the final latency data.

[0019] Optionally, based on the final latency data of each node, a minimum connected graph under a preset network traffic is obtained, including:

[0020] Based on the preset network traffic information and the final latency data of each node, the Prim algorithm is used to generate the minimum connected graph of the current cloud virtual network under specified termination conditions.

[0021] The preset network traffic information is the network traffic information after the cloud virtual network is actually deployed.

[0022] Optionally, based on the final delay data of each node, the shortest path between any two nodes in the minimum connected graph is traversed, and an improved variable neighborhood search algorithm is used to expand and search the neighborhood of the shortest path to obtain a network topology graph that satisfies the optimal delay, including:

[0023] The termination conditions of the improved variable neighborhood search algorithm include: neighborhood vibration operator, and termination conditions after fusing local search and search optimization optimal value.

[0024] The improved variable neighborhood search algorithm uses three neighborhood vibration operators to expand the neighborhood. The three neighborhood vibration operators include: node replacement, node deletion, and node addition.

[0025] For node replacement, if there is more than one node between two nodes, a node in E that does not appear in the minimum connected graph is randomly selected to replace one of the three nodes and form a new path.

[0026] For node deletion, if there are more than one node between two nodes, randomly delete one of the three nodes and form a new direct path connecting the remaining nodes.

[0027] For node addition, for two nodes that are directly connected in the minimum connected graph, randomly select a node in E that does not appear in the minimum connected graph, add it between the two nodes, and form a new path;

[0028] An improved variable neighborhood search algorithm is used to iterate the shortest path. New paths are formed by three neighborhood vibration operators to iteratively obtain the optimal time delay between each node and obtain the network topology graph with the optimal time delay. E is the set of edges of the cloud virtual network.

[0029] Optionally, based on the final delay data of each node, the shortest path between any two nodes in the minimum connected graph is traversed, and an improved variable neighborhood search algorithm is used to expand and search the neighborhood of the shortest path to obtain a network topology graph that satisfies the optimal delay. This also includes:

[0030] Given T = (V, D) and G = (V, E) as inputs to the improved variable neighborhood search algorithm, the improved variable neighborhood search algorithm outputs G. r =(V,E) r T is the minimum connected graph, G is the cloud virtual network, V is the set of all nodes in the minimum connected graph, D is the set of edges in the minimum connected graph, and E is the set of edges in the cloud virtual network. r The minimum time delay edge between all pairs of nodes in the output network topology graph, k is the neighborhood vibration operator, iter is the number of iterations, and t is the computation time;

[0031] Initial conditions: iter = 0, t = 0, E r=D;

[0032] In the iterative process, k=1, v1 and v2 select random nodes in V, and choose any line E' between them; when k≤k max At that time, the neighborhood solution E” of E’ is generated according to the k-th neighborhood vibration operator; if len(E”)>1, then the search for two lines E”' with the same physical nodes and the same len is repeated in E; if the time delay of E”' is less than E”, then E” = E”' is set until there is no E”' that meets the condition; if the time delay of the edge e” between two nodes in E is higher than E”, then k=1, E r =E r +E”, E r =E r -e”; otherwise, k = k + 1;

[0033] k = k + 1, repeat the above iterative process until t reaches the termination time or the iteration count iter reaches its maximum value, and obtain G. r .

[0034] Optionally, it also includes:

[0035] According to the output G r Obtain the degree of each node in the network topology graph, and then... r The degree of each node serves as auxiliary decision-making information when the cloud virtual network is actually implemented.

[0036] This invention also discloses a cloud virtual network node deployment auxiliary decision-making system, comprising:

[0037] The data simulation acquisition unit is used to acquire the forwarding latency data of each node in the full set of candidate nodes in the cloud network. The latency data is the forwarding data of each node under each network traffic when the cloud virtual network is online, which is simulated by the streaming system. All latency data of each node is cleaned, and the cleaned data under different time periods and different network traffic are weighted and fused to obtain the final latency data of each node that matches the actual scenario.

[0038] The network topology graph acquisition unit is used to acquire a minimum connected graph under a preset network traffic based on the final latency data of each node. The minimum connected graph is a star topology structure containing latency data within a specified range between any two nodes. Based on the final latency data of each node, the unit traverses the shortest path between any two nodes in the minimum connected graph and uses an improved variable neighborhood search algorithm to expand and search the neighborhood of the shortest path to obtain a network topology graph that satisfies the optimal latency and the degree of each node in the network topology graph.

[0039] This invention also discloses an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus;

[0040] The memory is used to store computer programs;

[0041] When the processor executes a program stored in the memory, it implements the method described in the embodiments of the present invention.

[0042] This invention also discloses one or more computer-readable media storing instructions that, when executed by one or more processors, cause the processors to perform the methods described in this invention.

[0043] The embodiments of the present invention have the following advantages:

[0044] The auxiliary decision-making method provided in this invention detects and preprocesses the latency data of each node in a cloud virtual network. Based on the preprocessed latency data, it obtains a minimum connected graph under a preset network traffic. Then, combining the latency between candidate physical nodes, it expands and searches the neighborhood of any node in the minimum connected graph using a variable neighborhood search algorithm, thereby obtaining the network topology graph with the shortest latency for each node. In other words, by optimizing the cloud virtual network design scheme, it can provide suggestions on the deployment carriers of different network element nodes, eliminating the subtle impact on network performance caused by deployment reasons in the prior art. At the same time, it eliminates the impact of high latency communication between network element nodes after the previous coarse deployment on the overall network performance.

[0045] The method of this invention solves the limitation of transmission latency between physical devices on traffic transmission latency on cloud virtual networks in the prior art. The method of this invention further optimizes the impact of traditional coarse-grained deployment on virtual network latency through a complete solution from simulation, monitoring, optimization calculation to deployment suggestions for network element nodes.

[0046] The decision support system of this invention can calculate a network topology optimization model for massive physical nodes, supported by an improved variable neighborhood search algorithm. For complex network topologies, it generates a topology graph containing routes with short latency between any two nodes, providing feasible, reliable, and efficient network element deployment suggestions. It also offers solution suggestions under different conditions due to factors such as actual network conditions, geographical environment, and data center environment. Furthermore, it fully balances the accuracy of the results with the computation time, making the entire system not only usable but also easy to use. Attached Figure Description

[0047] Figure 1This is a flowchart illustrating the steps of an auxiliary decision-making method for deploying cloud virtual network nodes provided in an embodiment of the present invention.

[0048] Figure 2 This is a flowchart of another auxiliary decision-making method for deploying cloud virtual network nodes provided in an embodiment of the present invention;

[0049] Figure 3 This is an example diagram of the minimum connected graph provided in the embodiments of the present invention;

[0050] Figure 4 This is an example diagram of physical node replacement as a neighborhood vibration operator in an embodiment of the present invention;

[0051] Figure 5 This is an example diagram of physical node deletion as a neighborhood vibration operator in an embodiment of the present invention;

[0052] Figure 6 This is an example diagram of physical nodes added as neighborhood vibration operators in an embodiment of the present invention;

[0053] Figure 7 This is an example diagram of the network topology with optimal latency in an embodiment of the present invention;

[0054] Figure 8 This is a structural block diagram of a cloud virtual network node deployment auxiliary decision-making system provided in an embodiment of the present invention;

[0055] Figure 9 This is a block diagram of an electronic device provided in an embodiment of the present invention;

[0056] Figure 10 This is a schematic diagram of a computer-readable medium provided in an embodiment of the present invention. Detailed Implementation

[0057] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0058] In this embodiment of the invention, the nodes that acquire latency data are nodes from the entire set of candidate nodes, and each node in the entire set of candidate nodes is a physical device that can potentially serve as a network element node.

[0059] Reference Figure 1 The diagram illustrates a flowchart of a cloud virtual network node deployment auxiliary decision-making method provided in an embodiment of the present invention, which may specifically include the following steps:

[0060] 101. Obtain the forwarding latency data of each node in the full set of candidate nodes in the cloud network. The latency data is the forwarding data of each node under each network traffic when the cloud virtual network is simulated to go online with the help of the streaming system.

[0061] In this embodiment, the first latency data between any two nodes at different time intervals can be obtained by using the streaming system to simulate a cloud virtual network without load; and the second latency data between any two nodes at different time intervals can be obtained by using the streaming system to simulate a cloud virtual network traffic of n specified network traffic; where n is greater than or equal to 1; at this time, the first latency data and the second latency data constitute the latency data forwarded by each node.

[0062] In this embodiment, the virtual network element is deployed on the physical node, while the main body for detecting latency in step 101 is the cloud network node (i.e., the physical node), at which point the virtual network element has not yet been deployed.

[0063] In this embodiment, when the cloud virtual network has no actual load, different network traffic is simulated by the traffic-drilling system to collect basic latency data. This helps to establish a performance benchmark for comparing network performance under different load conditions.

[0064] Gradually increase the network load to n specified traffic levels, each level should include different data transmission rates and types (e.g., TCP / UDP). At each load level, record latency data between any two nodes, which can help identify network bottlenecks and performance degradation points.

[0065] In addition to recording latency data at specific points in time, the average latency over a period of time should also be calculated to better understand the overall performance trend of the network. Furthermore, analyzing the distribution of latency data, such as statistical indicators like minimum, maximum, and median values, can reveal the volatility and stability of network latency.

[0066] 102. Clean all latency data for each node, and perform weighted fusion of the cleaned data under different time periods and different network traffic to obtain the final latency data for each node that matches the actual scenario.

[0067] For example, for all latency data of each node, density-based outlier detection is used to remove outliers from the latency data, and k-nearest neighbor interpolation is used to process noise in the latency data to obtain preprocessed latency data. Based on the actual network traffic after the cloud virtual network is deployed, the preprocessed latency data under different times and network traffic is weighted and fused to obtain the final latency data.

[0068] In this embodiment, cleaning and weighted fusion of the latency data of each node in the virtual network is a key step to ensure the accuracy and effectiveness of data analysis.

[0069] When performing data cleaning, density-based outlier detection is used to remove outliers from latency data. Specifically, this includes:

[0070] Density-based clustering algorithms such as DBSCAN (density clustering algorithm) are used to automatically identify and label outliers whose density differs significantly from that of other data points. This method is particularly suitable for processing abrupt changes in latency data, and it can effectively remove outliers caused by network failures, hardware problems, or other abnormal factors.

[0071] Noise processing in time-delay data using k-nearest neighbor interpolation includes:

[0072] For data points marked as outliers, the average of their k nearest neighbors is used to replace them. This method can remove local noise while preserving the overall trend of the data. However, choosing an appropriate value for k is crucial; typically, the optimal value needs to be determined experimentally to balance denoising effectiveness and data accuracy.

[0073] Furthermore, it is also necessary to handle missing data in the latency data:

[0074] For data missing in a short period of time, linear interpolation can be used to fill the gaps, which involves linear interpolation based on the time delay values ​​before and after the missing data.

[0075] For data gaps of a relatively long period, polynomial interpolation can be used to fit the curve using multiple surrounding data points, generating a smoother time delay curve.

[0076] In addition, time series models such as ARIMA (Autoregressive Integral Moving Average) can be used to predict the time delay value of the missing part based on historical data.

[0077] When performing weighted fusion, weights can be set based on time or network traffic. Specifically:

[0078] Latency data from different time periods may have different representativeness. For example, network traffic during working hours is usually higher than during non-working hours. Therefore, data from working hours can be given higher weight. In other words, a reasonable weight allocation strategy can be set according to actual business needs and network usage patterns.

[0079] Latency data under different network traffic conditions reflects network performance under varying loads. Different weights can be assigned to data under different traffic levels based on common traffic levels in real-world application scenarios. For example, if network traffic is at a moderate level most of the time in a real-world application scenario, higher weights can be given to latency data under moderate traffic.

[0080] 103. Based on the final latency data of each node, obtain the minimum connected graph under the preset network traffic, wherein the minimum connected graph is a star topology structure containing latency data within a specified range between any two nodes.

[0081] In the specific implementation process, this embodiment can generate the minimum connectivity graph of the current cloud virtual network using the Prim algorithm under specified termination conditions, based on preset network traffic information and the final latency data of each node. Here, the preset network traffic information refers to the network traffic information after the cloud virtual network is actually deployed.

[0082] Prim's algorithm is a classic algorithm for generating minimum spanning trees. It starts from an initial node and gradually adds edges until all nodes are included, forming a connected graph.

[0083] The specific steps of the algorithm are as follows:

[0084] Choose an initial node as the starting point;

[0085] Create a collection s, initially containing only the starting node;

[0086] Create a priority queue Q to store all nodes that have not been added to s and the minimum edge weight (delay) between them and nodes in s;

[0087] Remove the node u with the smallest edge weight from the priority queue Q and add it to the set s;

[0088] Update the edge weights of nodes adjacent to u in priority queue Q. If the delay of reaching an adjacent node v via u is less than the delay of the current record, then update the edge weight of v and its parent node.

[0089] Repeat the above steps until all nodes have been added to the set;

[0090] The algorithm terminates when all nodes have been added to set s, and the resulting graph is the minimum connected graph.

[0091] When selecting edges, ensure that the delay of newly added edges is within a preset range. If the delay of an edge exceeds the range, do not select that edge, and continue searching for other edges that meet the criteria.

[0092] When generating the minimum connected graph, pre-defined network traffic information is considered to ensure that the generated graph performs optimally under actual traffic conditions.

[0093] 104. Based on the final delay data of each node, traverse the shortest path between any two nodes in the minimum connected graph, and use an improved variable neighborhood search algorithm to expand and search the neighborhood of the shortest path to obtain the network topology graph that satisfies the optimal delay and the degree of each node in the network topology graph.

[0094] In this embodiment, the termination condition of the improved variable neighborhood search algorithm may include: neighborhood vibration operator, and termination condition after fusing local search and search optimization optimal value.

[0095] The improved variable neighborhood search algorithm uses three neighborhood vibration operators to expand the neighborhood. The three neighborhood vibration operators include: node replacement, node deletion, and node addition.

[0096] For node replacement, if there is more than one node between two nodes, a node in E that does not appear in the minimum connected graph is randomly selected to replace one of the three nodes and form a new path.

[0097] For node deletion, if there is more than one node between two nodes, randomly delete one of the three nodes and form a new path that directly connects the remaining nodes;

[0098] For node addition, for two nodes that are directly connected in the minimum connected graph, randomly select a node in E that does not appear in the minimum connected graph, add it between the two nodes, and form a new path;

[0099] For example, given a selected pair of nodes A and B, if there is at least one other station between them, a node C is randomly selected from the set of nodes E that is not in the current connected graph, and C is used to replace a node between A and B to form a new path.

[0100] Similarly, for a pair of nodes A and B, if there is at least one other node between them, one of the nodes D is randomly selected and deleted, so that A and B are directly connected, forming a shorter path.

[0101] For two directly connected nodes A and B, a node F is randomly selected from the set of nodes E that are not in the current connected graph and inserted between A and B. This increases the complexity of the network and can also improve latency performance.

[0102] An improved variable neighborhood search algorithm is used to iterate the shortest path. New paths are formed by three neighborhood vibration operators to iteratively obtain the optimal time delay between each node and obtain the network topology graph with the optimal time delay. E is the set of edges of the cloud virtual network.

[0103] Specifically, at the beginning of each iteration, one of the three defined neighborhood vibration operators is selected for application, either in a fixed order or randomly.

[0104] Perform the corresponding operation (replace, delete, or add nodes) based on the selected neighborhood vibration operator to generate candidate solutions; calculate the total time delay of the newly generated network topology graph and compare it with the current best solution;

[0105] If a candidate solution is better than the current best solution, it is accepted as the new current solution, and the search continues. Otherwise, a decision is made based on certain acceptance criteria to decide whether to accept a worse solution in order to avoid getting trapped in local optima.

[0106] The algorithm terminates when the preset maximum number of iterations is reached, when the current solution is not improved after a number of consecutive iterations, or when all possible neighborhoods have been explored and no better solution has been found.

[0107] Specifically, when the neighborhood vibration operator can no longer produce effective changes (i.e., it is impossible to form a new path by replacing, deleting, or adding nodes), the search space is considered to have been fully explored.

[0108] The final network topology graph should satisfy the communication needs between all nodes and have optimal latency performance. Then, calculate and output the degree of each node in the network topology graph, which is the number of edges connected to each node.

[0109] To facilitate understanding of step 104 above, this step can be explained in detail as follows:

[0110] Given T = (V, D) and G = (V, E) as inputs to the improved variable neighborhood search algorithm, the improved variable neighborhood search algorithm outputs G. r =(V,E) r T is the minimum connected graph, G is the cloud virtual network, V is the set of all nodes in the minimum connected graph, D is the set of edges in the minimum connected graph, and E is the set of edges in the cloud virtual network. r The minimum time delay edge between all pairs of nodes in the output network topology graph, k is the neighborhood vibration operator, iter is the number of iterations, and t is the computation time;

[0111] Initial conditions: iter = 0, t = 0, E r =D;

[0112] In the iterative process, k=1, v1 and v2 select random nodes in V, and choose any line E' between them; when k≤k max At that time, the neighborhood solution E” of E’ is generated according to the k-th neighborhood vibration operator; if len(E”)>1, then the search for two lines E”' with the same physical nodes and the same len is repeated in E; if the time delay of E”' is less than E”, then E” = E”' is set until there is no E”' that meets the condition; if the time delay of the edge e” between two nodes in E is higher than E”, then k=1, E r =E r +E”, E r =E r -e”; otherwise, k = k + 1;

[0113] k = k + 1, repeat the above iterative process until t reaches the termination time or the iteration count iter reaches its maximum value, and obtain G. r .

[0114] Furthermore, in this embodiment, based on the output G r Obtain the degree of each node in the network topology graph, and then... r The degree of each node serves as auxiliary decision-making information when the cloud virtual network is actually implemented.

[0115] The method in this embodiment detects and preprocesses the latency data of each node in the cloud virtual network. Based on the preprocessed latency data, it obtains the minimum connected graph under a preset network traffic. Then, combining the latency between candidate physical nodes, it expands and searches the neighborhood of any node in the minimum connected graph using a variable neighborhood search algorithm, thereby obtaining the network topology graph with the shortest latency for each node. In other words, by optimizing the cloud virtual network design scheme, it can provide suggestions for the deployment carriers of different network element nodes, eliminating the subtle impact on network performance caused by deployment reasons in the prior art. At the same time, it eliminates the impact of high latency communication between network element nodes after the previous coarse deployment on the overall network performance.

[0116] In other words, the above method solves the limitation of transmission latency between physical devices on traffic transmission latency on cloud virtual networks in the prior art. The method of this embodiment further optimizes the impact of traditional rough and tumble deployment on virtual network latency through a complete solution from simulation, monitoring, optimization calculation to deployment suggestions for network element nodes.

[0117] In particular, the method in this embodiment is adaptable to traffic simulation methods across multiple scenarios. By leveraging multi-round data collection and data noise optimization methods tailored to the characteristics of physical networks, it provides a well-prepared traffic simulation model that fits the specific application scenario for the entire optimization process. It also offers a feasible solution for other research on network traffic simulation and detection.

[0118] like Figure 2 As shown, Figure 2 A flowchart of another decision-making aid method for deploying cloud virtual network nodes is shown, and the steps are explained below:

[0119] 201. Simulate the actual traffic data of each node in the full set of candidate nodes in the cloud virtual network, and obtain the latency data of each node collected in multiple rounds under multiple scenarios.

[0120] In this embodiment, in order to realize the forwarding capability between physical nodes under the network traffic scale of the simulated cloud virtual network, the latency between each node of the physical network is collected through network traffic simulation and multi-scenario multi-round collection, so as to provide accurate latency data for the subsequent network topology optimization processing stage, and at the same time, it can eliminate the pollution of data due to occasional vibrations and planned disturbances of the physical network during the collection.

[0121] Network traffic simulation and multi-scenario, multi-round data collection are combined to ensure that the collected data fully describes the actual network conditions. First, the idle time of the cloud virtual network (hereinafter referred to as the network) is simulated, i.e., the latency between any two network element nodes is tested and recorded when the network is unloaded. Second, data collection is performed during network busy times. Based on a preset traffic scale, n pre-set flow scales are set up to perform flow distribution between physical networks. By monitoring the physical switching equipment of the flow distribution system, the flow scale at physical nodes is dynamically adjusted, thereby adjusting the overall traffic scale within the network. Network latency is collected when the expected traffic scale is reached. Third, to avoid regular interference caused by periodic probing, probing is performed at uncertain time intervals, while continuous long-term monitoring is conducted to collect network latency data over a long period to observe the trend of network performance changes over time.

[0122] When probing at uncertain time intervals, it is also necessary to clean the collected data and identify and eliminate data anomalies caused by factors such as occasional network fluctuations and planned disturbances.

[0123] Specifically, simulating network idle time includes: selecting any two nodes in the network for latency testing under no additional load, recording the latency data between different node pairs, mainly maintaining the consistency of test conditions, such as test time period and test frequency; analyzing this latency data to understand the basic latency characteristics of the network under no-load conditions.

[0124] 202. Data cleaning.

[0125] To address missing values ​​and anomalies in the latency data of each node obtained in step 201, optimizations are implemented to adapt to specific application scenarios.

[0126] In this embodiment, a density-based anomaly detection method is used for anomaly detection. At the same time, the human factors such as network maintenance can be eliminated by the screening of deployment personnel. Considering the characteristics of network stability, the k-nearest neighbor interpolation method is used for noise processing.

[0127] 203. Data fusion.

[0128] In this embodiment, data collected at different times and with different traffic volumes can be weighted and fused according to the actual network traffic volume after the network is deployed, so as to provide data that fits the actual scenario for subsequent optimization.

[0129] For example, if the bandwidth of a physical node is 10M under actual network traffic scale, latency was collected / probeed at 1M, 2M, 3M, 4M, 5M, 6M, 7M, 8M, 9M, and 10M in multiple rounds and scenarios. Then, before planning network elements, if it is determined that the expected load is likely to be 80% in most scenarios, the latency data probed at 8M is given a higher weight, such as 40%, while the latency data with bandwidth around 8M can have a lower weight, such as 10%, and the others are not weighted, to calculate an average. The weights for the aforementioned latency data are: 1M 20%, 2M 15%, 3M 15%, 4M 10%, 5M 10%, 6M 10%, 7M 6%, 8M 5%, 9M 5%, and 10M 4%. The above is merely an example of data fusion, and this embodiment is not limited to it. Adjustments can be made according to actual needs.

[0130] 204. Network topology optimization.

[0131] To achieve a star topology graph that allows communication between any two nodes within a given network traffic volume and includes low-latency connections between any two nodes, Prim's algorithm is used to generate a minimum connected graph of the current network, such as... Figure 3 As shown, the shortest path between any two physical nodes is then traversed. Based on the improved variable neighborhood search algorithm, the neighborhood of the shortest path is expanded and searched within the neighborhood to find a better solution between any two physical nodes. The computation time and result reliability of the optimization algorithm are balanced by using the neighborhood vibration factor and the termination condition after fusing local search and search optimization optimal values, thus obtaining a star topology map, as shown. Figure 7 As shown.

[0132] In this embodiment, the improved variable neighborhood search algorithm introduces three neighborhood vibration operators (such as physical node deletion, physical node replacement, and physical node addition) to expand the neighborhood. This prevents the algorithm from getting stuck in local conditions in the calculation of the minimum delay route between any two points, thus avoiding the inability to obtain a more optimized route. It also takes into account the computation time.

[0133] Physical node replacement occurs when there is more than one node between two nodes (e.g., edge network element nodes). Figure 4(a) As shown, PM1, PM2, and PM3 are used to form a new path. A node PM4 is randomly selected from the physical nodes of the non-minimum connected graph. This means a physical node in node E (excluding existing route points) is randomly selected to replace node 3 (originally PM2). Direct edges between the replaced node and the original node are then found to form a new route. Figure 4 (b) is shown; otherwise, no action is taken.

[0134] Physical node deletion occurs if there is more than one node between two edge nodes, such as... Figure 5 (a) Given PM1, PM2, and PM3, randomly select a node (PM2) to delete, and replace the edges between the nodes on either side of the deleted node with direct connections, such as... Figure 5 (b) is shown; otherwise, no action is taken.

[0135] Physical nodes are added, currently as follows: Figure 6 (a) shows the addition of a physical node (PM4) from E (excluding existing route points) to the route between two nodes (PM1, PM2), and the addition of direct edges between the new node and the existing nodes to form a new route, as shown. Figure 6 As shown in (b).

[0136] 205. Recommended network element nodes.

[0137] To translate the optimized network topology into recommendations for deploying network element nodes, a network is constructed based on the minimum latency path between any two nodes obtained from the network topology. At this point, the topology is no longer a minimum connected graph, but rather a network containing nodes where not all physical nodes have a degree of 1, such as... Figure 7 As shown, a node with a higher degree indicates that it has more edges with lower latency connecting to other nodes, making it more suitable as a network element node.

[0138] Specifically, based on the optimized network topology, it can be determined which nodes are more suitable for deployment as network element nodes. These nodes typically have high degrees and more connections with other nodes, and these connections have lower latency. The specific steps and methods are as follows:

[0139] ①Analyze the optimized network topology:

[0140] Calculate the degree of each node: The degree of a node is the number of edges directly connected to that node. In the optimized network topology, nodes with higher degrees typically have more low-latency connections.

[0141] Calculate the average latency of each node: For each node, calculate the average latency of the shortest path between it and all other nodes to evaluate the importance and performance of the node in the entire network.

[0142] ② Identify key nodes:

[0143] Select nodes with higher degree: Set a degree threshold and select nodes with a degree greater than or equal to the threshold as candidate network element nodes; sort the nodes from highest to lowest degree and select the top N nodes as preliminary candidate nodes.

[0144] Taking average latency into account: Set a threshold for average latency, and select nodes with average latency below that threshold; combine the node's degree and average latency to assign a score to each node. For example, a weighted sum method can be used, where degree has a higher weight and average latency has a lower weight.

[0145] The method in this embodiment can provide suggestions for deployment decisions of various forwarding devices in physical networks and offer solutions for optimizing lines between different regions of physical networks. It provides feasible solutions for other research on network traffic simulation and detection. Simultaneously, it optimizes the latency of physical network transmission between different regions, providing new ideas for network traffic simulation, monitoring, and collection in cloud service systems.

[0146] It should be noted that, for the sake of simplicity, the method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments of the present invention are not limited to the described order of actions, because according to the embodiments of the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions involved are not necessarily essential to the embodiments of the present invention.

[0147] Reference Figure 8 The diagram illustrates a structural block diagram of a cloud virtual network node deployment auxiliary decision-making system provided in an embodiment of the present invention, which may specifically include the following units:

[0148] The data simulation acquisition unit is used to acquire the forwarding latency data of each node in the full set of candidate nodes in the cloud network. The latency data is the forwarding data of each node under each network traffic when the cloud virtual network is online, which is simulated by the streaming system. All latency data of each node is cleaned, and the cleaned data under different time periods and different network traffic are weighted and fused to obtain the final latency data of each node that matches the actual scenario.

[0149] The network topology graph acquisition unit is used to acquire a minimum connected graph under a preset network traffic based on the final latency data of each node. The minimum connected graph is a star topology structure containing latency data within a specified range between any two nodes. Based on the final latency data of each node, the unit traverses the shortest path between any two nodes in the minimum connected graph and uses an improved variable neighborhood search algorithm to expand and search the neighborhood of the shortest path to obtain a network topology graph that satisfies the optimal latency and the degree of each node in the network topology graph.

[0150] The decision support system in this embodiment can compute a network topology optimization model for a massive number of physical nodes, supported by an improved variable neighborhood search algorithm. For complex network topologies, it generates a topology graph containing routes with short latency between any two nodes, providing feasible, reliable, and efficient network element deployment suggestions. It also offers solution suggestions under different conditions due to factors such as actual network conditions, geographical environment, and data center environment. Furthermore, it fully balances the accuracy of the results with the computation time, making the entire system not only usable but also easy to use.

[0151] As the device embodiment is basically similar to the method embodiment, the description is relatively simple, and relevant parts can be found in the description of the method embodiment.

[0152] In addition, embodiments of the present invention also provide an electronic device, such as... Figure 9 As shown, it includes a processor 1301, a communication interface 1302, a memory 1303, and a communication bus 1304. The processor 1301, the communication interface 1302, and the memory 1303 communicate with each other through the communication bus 1304.

[0153] Memory 1303 is used to store computer programs;

[0154] When processor 1301 executes a program stored in memory 1303, it performs the following steps:

[0155] Acquire the latency data of each node forwarding in the cloud network. The latency data is the data of each node forwarding under each network traffic when the cloud virtual network is simulated to go online with the help of the streaming system.

[0156] All latency data for each node is cleaned, and the cleaned data from different time periods and under different network traffic conditions are weighted and fused to obtain the final latency data for each node that matches the actual scenario.

[0157] Based on the final latency data of each node, obtain the minimum connectivity graph under the preset network traffic. The minimum connectivity graph is a star topology structure that contains latency data within a specified range between any two nodes.

[0158] Based on the final latency data of each node, the shortest path between any two nodes in the minimum connected graph is traversed. An improved variable neighborhood search algorithm is used to expand and search the neighborhood of the shortest path to obtain a network topology graph that satisfies the optimal latency and the degree of each node in the network topology graph.

[0159] The communication bus mentioned above can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used to represent it in the diagram, but this does not mean that there is only one bus or one type of bus.

[0160] The communication interface is used for communication between the aforementioned terminal and other devices.

[0161] The memory may include random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0162] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0163] like Figure 10 As shown, in another embodiment of the present invention, a computer-readable storage medium 1401 is also provided, which stores instructions that, when executed on a computer, cause the computer to execute the auxiliary decision-making method for deploying cloud virtual network nodes as described in the above embodiment.

[0164] In another embodiment of the present invention, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to execute the auxiliary decision-making method for deploying cloud virtual network nodes as described in the above embodiments.

[0165] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid state disk (SSD)).

[0166] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0167] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0168] 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 are included within the scope of protection of the present invention.

Claims

1. A method for assisting decision-making in the deployment of cloud virtual network nodes, characterized in that, include: Obtain the forwarding latency data of each node in the full set of candidate nodes in the cloud network. The latency data is the forwarding data of each node under each network traffic when the cloud virtual network goes online, which is simulated by the traffic blocking system. All latency data for each node is cleaned, and the cleaned data from different time periods and under different network traffic conditions are weighted and fused to obtain the final latency data for each node that matches the actual scenario. Based on the final latency data of each node, obtain the minimum connectivity graph under the preset network traffic. The minimum connectivity graph is a star topology structure that contains latency data within a specified range between any two nodes. Based on the final delay data of each node, the shortest path between any two nodes in the minimum connected graph is traversed. An improved variable neighborhood search algorithm is used to expand and search the neighborhood of the shortest path to obtain a network topology graph that satisfies the optimal delay and the degree of each node in the network topology graph. The process involves traversing the shortest path between any two nodes in the minimum connected graph based on the final latency data of each node, and using an improved variable neighborhood search algorithm to expand and search the neighborhood of the shortest path to obtain a network topology that satisfies the optimal latency. This includes: The termination conditions of the improved variable neighborhood search algorithm include: neighborhood vibration operator, and termination conditions after fusing local search and search optimization optimal value. The improved variable neighborhood search algorithm uses three neighborhood vibration operators to expand the neighborhood. The three neighborhood vibration operators include: node replacement, node deletion, and node addition. Given T = (V, D) and G = (V, E) as inputs to the improved variable neighborhood search algorithm, the improved variable neighborhood search algorithm outputs G. r =(V,E) r T is the minimum connected graph, G is the cloud virtual network, V is the set of all nodes in the minimum connected graph, D is the set of edges in the minimum connected graph, and E is the set of edges in the cloud virtual network. r The minimum time delay edge between all pairs of nodes in the output network topology graph, k is the neighborhood vibration operator, iter is the number of iterations, and t is the computation time; Initial conditions: iter = 0, t = 0, E r =D; In the iterative process, k=1, v1 and v2 select random nodes in V, and choose any line E' between them; when k≤k max At that time, the neighborhood solution E” of E’ is generated according to the k-th neighborhood vibration operator; if len(E”)>1, then the search for two lines E”' with the same physical nodes and the same len is repeated in E; if the time delay of E”' is less than E”, then E” = E”' is set until there is no E”' that meets the condition; if the time delay of the edge e” between two nodes in E is higher than E”, then k=1, E r =E r +E”, E r =E r -e”; otherwise, k = k + 1; k = k + 1, repeat the above iterative process until t reaches the termination time or the iteration count iter reaches its maximum value, and obtain G. r .

2. The method according to claim 1, characterized in that, Obtain the forwarding latency data for each node in the entire set of candidate nodes in the cloud network, including: By using a streaming system to simulate a cloud virtual network without load, the first latency data between any two nodes at different time intervals is obtained; When the virtual network traffic of the streaming system reaches n specified network traffic values, the second latency data between any two nodes at different time intervals is obtained; n is greater than or equal to 1. The first delay data and the second delay data constitute the delay data forwarded by each node.

3. The method according to claim 1, characterized in that, All latency data for each node is cleaned, and the cleaned data from different time periods and under different network traffic conditions are weighted and fused, including: For all latency data of each node, a density-based outlier detection method is used to remove outliers from the latency data, and a k-nearest neighbor interpolation method is used to process noise in the latency data to obtain preprocessed latency data. Based on the actual network traffic after the cloud virtual network is deployed, the preprocessed latency data under different times and network traffic conditions are weighted and merged to obtain the final latency data.

4. The method according to claim 1, characterized in that, Based on the final latency data of each node, obtain the minimum connectivity graph under the preset network traffic, including: Based on the preset network traffic information and the final latency data of each node, the Prim algorithm is used to generate the minimum connected graph of the current cloud virtual network under specified termination conditions. The preset network traffic information is the network traffic information after the cloud virtual network is actually deployed.

5. The method according to claim 1, characterized in that, Based on the final latency data of each node, the shortest path between any two nodes in the minimum connected graph is traversed. An improved variable neighborhood search algorithm is used to expand and search the neighborhood of the shortest path to obtain a network topology graph that satisfies the optimal latency. This also includes: For node replacement, if there is more than one node between two nodes, a node in E that does not appear in the minimum connected graph is randomly selected to replace one of the three nodes and form a new path. For node deletion, if there are more than one node between two nodes, randomly delete one of the three nodes and form a new direct path connecting the remaining nodes. For node addition, for two nodes that are directly connected in the minimum connected graph, randomly select a node in E that does not appear in the minimum connected graph, add it between the two nodes, and form a new path; An improved variable neighborhood search algorithm is used to iterate the shortest path. New paths are formed by three neighborhood vibration operators to iteratively obtain the optimal time delay between each node and obtain the network topology graph with the optimal time delay. E is the set of edges of the cloud virtual network.

6. The method according to claim 1, characterized in that, Also includes: According to the output G r Obtain the degree of each node in the network topology graph, and then... r The degree of each node serves as auxiliary decision-making information when the cloud virtual network is actually implemented.

7. A cloud virtual network node deployment auxiliary decision-making system, characterized in that, include: The data simulation acquisition unit is used to acquire the forwarding latency data of each node in the full set of candidate nodes in the cloud network. The latency data is the forwarding data of each node under each network traffic when the cloud virtual network is online, which is simulated by the streaming system. All latency data of each node is cleaned, and the cleaned data under different time periods and different network traffic are weighted and fused to obtain the final latency data of each node that matches the actual scenario. The network topology graph acquisition unit is used to acquire the minimum connected graph under the preset network traffic based on the final latency data of each node. The minimum connected graph is a star topology structure containing latency data within a specified range between any two nodes. Based on the final latency data of each node, the unit traverses the shortest path between any two nodes in the minimum connected graph and uses an improved variable neighborhood search algorithm to expand and search the neighborhood of the shortest path to obtain the network topology graph that satisfies the optimal latency and the degree of each node in the network topology graph. The termination conditions of the improved variable neighborhood search algorithm include: a neighborhood vibration operator and a termination condition resulting from the fusion of local search and search optimization optimal values. The three neighborhood vibration operators in the improved variable neighborhood search algorithm are used to expand the neighborhood. These three operators include: node replacement, node deletion, and node addition. They are also used to input T = (V, D) and G = (V, E) into the improved variable neighborhood search algorithm, and the improved variable neighborhood search algorithm outputs G. r =(V,E) r T is the minimum connected graph, G is the cloud virtual network, V is the set of all nodes in the minimum connected graph, D is the set of edges in the minimum connected graph, and E is the set of edges in the cloud virtual network. r The minimum time delay edge between all pairs of nodes in the output network topology graph, k is the neighborhood vibration operator, iter is the number of iterations, and t is the computation time; Initial conditions: iter = 0, t = 0, E r =D; In the iterative process, k=1, v1 and v2 select random nodes in V, and choose any line E' between them; when k≤k max At that time, the neighborhood solution E” of E’ is generated according to the k-th neighborhood vibration operator; if len(E”)>1, then the search for two lines E”' with the same physical nodes and the same len is repeated in E; if the time delay of E”' is less than E”, then E” = E”' is set until there is no E”' that meets the condition; if the time delay of the edge e” between two nodes in E is higher than E”, then k=1, E r =E r +E”, E r =E r -e”; otherwise, k = k + 1; k = k + 1, repeat the above iterative process until t reaches the termination time or the iteration count iter reaches its maximum value, and obtain G. r .

8. An electronic device, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; The memory is used to store computer programs; When the processor executes a program stored in the memory, it implements the method as described in any one of claims 1-5.

9. A computer-readable medium having instructions stored thereon that, when executed by one or more processors, cause the processors to perform the method as described in any one of claims 1-5.

Images