A cloud computing network data transmission performance index evaluation system and method

Abstract

The application provides a cloud computing network data transmission performance index identification and evaluation processing system and method, wherein the method comprises the following steps: collecting transmission data in nodes of each transmission path of the cloud computing network; preprocessing the transmission data to obtain multidimensional performance index data; dividing the multidimensional performance index data into period samples, and constructing an index change matrix for each period sample; correcting and calculating the index change matrix by using a correction weight coefficient set to obtain a node load correction score; and screening and optimizing a transmission path by using the node load correction score of the node of each transmission path to form a path selection strategy. The above method improves the utilization rate of network resources and ensures efficient data transmission in the cloud computing environment.

Description

Technical Field

[0001] This invention relates to transmission evaluation metrics, and more particularly to a system and method for evaluating the performance indicators of cloud computing network data transmission. Background Technology

[0002] With the rapid development of cloud computing technology, more and more enterprises and individuals rely on cloud computing platforms for data storage, processing, and applications. Cloud computing, by providing flexible, efficient, and scalable computing resources, has been widely adopted in various business scenarios. However, performance evaluation of cloud computing networks has always been a crucial aspect of achieving efficient network management and optimization. Especially in complex network environments, factors such as data transmission latency, packet loss rate, bandwidth utilization, and node load can all affect the overall network performance.

[0003] Traditional network performance evaluation methods typically use static and simplified metrics, which fail to fully reflect the dynamic and time-varying characteristics of cloud computing networks. Transmission paths within a network can be affected by various factors, such as traffic variations, node load fluctuations, and changes in network topology, all of which cause transmission performance to change over time. Therefore, how to evaluate and optimize network performance in real-time and dynamically has become a pressing issue in cloud computing networks.

[0004] To address these issues, traditional network evaluation methods typically focus on a single performance metric, lacking comprehensive multi-dimensional and multi-time-period analysis, and failing to fully consider changes in network status across different time periods. As cloud computing networks become increasingly complex and large-scale, relying solely on traditional evaluation methods is insufficient to meet the demands for accurate network performance assessment and real-time optimization. Summary of the Invention

[0005] The purpose of this invention is to provide a system and method for evaluating the performance indicators of cloud computing network data transmission, which solves the above-mentioned technical problems pointed out in the prior art.

[0006] This invention provides a cloud computing network data transmission performance indicator identification and evaluation processing system, including a data center sampling and analysis module and a node load analysis and correction processing module:

[0007] The data center sampling and analysis module is used to collect transmission data from nodes along various transmission paths in the cloud computing network; preprocess the transmission data to obtain multi-dimensional performance index data; the multi-dimensional performance index data includes transmission delay, packet loss rate, bandwidth utilization, and node load data;

[0008] The node load analysis and correction processing module is used to divide the multidimensional performance index data into time period samples, construct an index change matrix for each time period sample, record the collection timestamp of the multidimensional performance index data, analyze the node load data of the multidimensional performance index data at each timestamp, use the node load data in each time period sample to introduce a time decay coefficient for calculation, and obtain a set of correction weight coefficients; use the set of correction weight coefficients to correct the index change matrix to obtain a node load correction score; and use the node load correction scores of the nodes in each transmission path to select the preferred transmission path, thus forming a path selection strategy.

[0009] Accordingly, this invention also proposes a method for evaluating and assessing the performance indicators of cloud computing network data transmission, comprising the following steps:

[0010] The transmission data of each node in the cloud computing network is collected; the transmission data is preprocessed to obtain multi-dimensional performance index data; the multi-dimensional performance index data includes transmission delay, packet loss rate, bandwidth utilization rate and node load data.

[0011] The multidimensional performance index data is divided into time period samples, and an index change matrix is ​​constructed for each time period sample. The multidimensional performance index data is recorded and collected with timestamps. The node load data of the multidimensional performance index data at each timestamp is analyzed. A time decay coefficient is introduced using the node load data in each time period sample to calculate and obtain a set of corrected weight coefficients. The index change matrix is ​​corrected using the set of corrected weight coefficients to obtain a node load correction score. The node load correction scores of the nodes in each transmission path are used to select the preferred transmission path to form a path selection strategy.

[0012] Compared with the prior art, the embodiments of the present invention have at least the following technical advantages:

[0013] Analysis of the cloud computing network data transmission performance evaluation system and method provided by the present invention shows that, in specific applications, the system first collects real-time data from each node in the cloud computing network through the data center sampling and analysis module 21, and preprocesses the transmitted data to extract multi-dimensional performance indicator data (such as transmission delay, packet loss rate, bandwidth utilization, and node load data). The method of collecting multi-dimensional performance indicator data can comprehensively reflect the network's performance in different aspects, providing a more accurate and comprehensive basis for subsequent evaluation and optimization.

[0014] Furthermore, the node load analysis and correction processing module can effectively and dynamically adjust the load assessment results by dividing the samples into time periods, constructing an index change matrix, and combining it with a time decay coefficient to correct the node load data, thus avoiding the error and lag problems in traditional static assessment methods. In particular, the introduction of the time decay coefficient enables the system to respond more accurately to changes in network load, thereby providing a reliable basis for dynamically adjusting network paths.

[0015] Furthermore, the system also uses the corrected node load score to select the best transmission path and optimize the path selection strategy; it can intelligently select paths with lighter load and better performance, thereby avoiding network congestion and improving the efficiency and stability of data transmission; by dynamically adjusting the path selection, the system can adapt to network load fluctuations in real time, further improving the overall performance and reliability of the network.

[0016] Furthermore, another significant advantage of this system lies in its automation and intelligence, which avoids manual intervention and outdated static evaluation methods; it can make real-time path optimization based on real-time network conditions, reducing network latency and packet loss rate, improving the utilization of network resources, and ensuring efficient data transmission in cloud computing environments. Attached Figure Description

[0017] Figure 1 This is an overall operation flowchart of a cloud computing network data transmission performance index identification and evaluation processing system according to Embodiment 1.

[0018] Figure 2 This is a flowchart of a cloud computing network data transmission performance index identification and evaluation processing system according to Embodiment 1;

[0019] Figure 3 This is a flowchart of a cloud computing network data transmission performance index identification and evaluation method according to Example 2;

[0020] Figure 4 This is a flowchart illustrating a method for evaluating and assessing cloud computing network data transmission performance indicators, specifically a method for determining path selection strategies using node load correction scoring, as described in Example 2.

[0021] Figure 5 This is a flowchart illustrating the process of obtaining the final path selection strategy through an initial path selection strategy in a cloud computing network data transmission performance index identification and evaluation method according to Embodiment 2.

[0022] Figure 6 This is a flowchart illustrating the path selection strategy reconstruction process in a cloud computing network data transmission performance index identification and evaluation method according to Example 2.

[0023] Figure 7This is a schematic diagram of a data stream copy for a cloud computing network data transmission performance index identification and evaluation processing method according to Embodiment 2;

[0024] Labels: Cloud computing network 100; physical host 10; virtual machine 11; routing device 12; load balancer 13; evaluation and processing system 200; data center sampling and analysis module 21; node load analysis and correction processing module 22. Detailed Implementation

[0025] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings.

[0027] Example 1

[0028] like Figure 1 as well as Figure 2 As shown, this embodiment of the invention provides a cloud computing network data transmission performance index identification and evaluation processing system (hereinafter referred to as the evaluation processing system) 200, including a data center sampling and analysis module 21 and a node load analysis and correction processing module 22:

[0029] The aforementioned evaluation and processing system 200 is used to evaluate the transmission path between multiple nodes on the cloud computing network 100, which is generally composed of physical hosts 10, virtual machines 11 (i.e., nodes), routing devices 12 (or network switching devices), load balancers, etc.

[0030] Physical hosts are the infrastructure of cloud computing, providing computing, storage, and network resources. Virtual machines (VMs), as virtualized logical nodes, run on physical hosts and are the basic units for actual task processing in cloud computing. Routing devices are responsible for forwarding network traffic and selecting paths, ensuring efficient data transmission between nodes. Load balancers distribute traffic across multiple nodes to avoid single-point overload. Management nodes, such as the Controller node in OpenStack and the Master node in Kubernetes, are responsible for resource scheduling and network policy management.

[0031] In summary, this cloud computing network generally consists of physical hosts 10, virtual machines 11 (i.e., nodes), routing devices 12, load balancers 13, and management nodes.

[0032] The data center sampling and analysis module 21 is used to collect transmission data from nodes along various transmission paths in the cloud computing network; preprocess the transmission data to obtain multi-dimensional performance index data; the multi-dimensional performance index data includes transmission delay, packet loss rate, bandwidth utilization rate, and node load data.

[0033] The node load analysis and correction processing module 22 is used to divide the multidimensional performance index data into time period samples, construct an index change matrix for each time period sample, record the collection timestamp of the multidimensional performance index data, analyze the node load data of the multidimensional performance index data at each timestamp, use the node load data in each time period sample to introduce a time decay coefficient for calculation, and obtain a set of correction weight coefficients; use the set of correction weight coefficients to correct the index change matrix to obtain a node load correction score; and use the node load correction scores of the nodes in each transmission path to select the preferred transmission path, thus forming a path selection strategy.

[0034] Example 2

[0035] like Figure 3 As shown in the figure, this embodiment of the invention provides a method for evaluating and assessing the performance indicators of cloud computing network data transmission, including the following steps:

[0036] S1: Collect transmission data from nodes along various transmission paths in the cloud computing network; preprocess the transmission data to obtain multi-dimensional performance index data;

[0037] The multidimensional performance metrics data include transmission delay, packet loss rate, bandwidth utilization, and node load data;

[0038] S2: Divide the multidimensional performance index data into time period samples, and construct an index change matrix for each time period sample; record the collection timestamps of the multidimensional performance index data, analyze the node load data of the multidimensional performance index data at each timestamp, and use the node load data in each time period sample to introduce a time decay coefficient for calculation to obtain a set of corrected weight coefficients; use the set of corrected weight coefficients to correct the index change matrix to obtain a node load correction score; use the node load correction scores of the nodes in each transmission path to select the preferred transmission path, thus forming a path selection strategy.

[0039] It should be noted that the above scheme optimizes multiple aspects of network performance. From data collection to path selection, each operation is designed to more accurately reflect the real-time status and load changes of the network. Furthermore, by dynamically correcting node load and introducing a time decay coefficient, the scheme ensures that path selection is more intelligent and can respond to changes in network load in real time. Its significance lies in improving the performance, stability, and adaptability of the network, thereby enabling the evaluation of network transmission performance.

[0040] Specifically, such as Figure 4 As shown, in step S2, the multidimensional performance index data is divided into time period samples, and an index change matrix is ​​constructed for each time period sample; the multidimensional performance index data is recorded and collected with time stamps, and the node load data of each time stamp is analyzed. A time decay coefficient is introduced using the node load data within each time period sample to calculate a set of corrected weight coefficients; the corrected weight coefficient set is used to correct the index change matrix to obtain a node load correction score; the node load correction score of each transmission path is used to select the preferred transmission path, thus forming a path selection strategy. The specific operation steps are as follows:

[0041] In steps S21-S23, the process of applying a sliding window to the multidimensional performance index data to obtain multiple time period samples is mainly described; and the index change matrix is ​​constructed using the multiple time period samples.

[0042] The index change matrix is ​​normalized, and the information entropy of each unit index is calculated. The minimum information entropy is selected, and the initial weight coefficient of each unit index in each row and column of the index change matrix is ​​calculated using the inverse information entropy principle to obtain the initial weight coefficient set.

[0043] The multidimensional performance index data is recorded and collected with timestamps. For each timestamp, the multidimensional performance index data is analyzed using node load data, and the fluctuation coefficient of the node is calculated. The fluctuation coefficient and the initial weight coefficient set are used to introduce a time decay coefficient for calculation to obtain the corrected weight coefficient set.

[0044] S21: Set a fixed time window for the multidimensional performance index data and slide the window to divide the multidimensional performance index data into segments, with each segment of multidimensional performance index data serving as a time period sample.

[0045] The dynamic change rate is calculated using multidimensional performance index data from samples in adjacent time periods;

[0046] For the dynamic change rate of multidimensional performance index data of samples in each time period, construct an index change matrix (i.e., the time window in which the behavior of the matrix changes over time, listed as different performance indicators (i.e., transmission delay, packet loss rate, bandwidth utilization, and node load data)).

[0047] It should be noted that the multidimensional performance index data is divided into multidimensional performance index data of n1, n2, n3, ..., n time period samples. The dynamic change rate is obtained by subtracting the multidimensional performance index data of n1 time period samples from the multidimensional performance index data of n2 time period samples, and then dividing by the multidimensional performance index data of n1 time period samples.

[0048] S22: Normalize the index change matrix and calculate the probability value using the normalized value of each unit index (i.e., the transmission delay, packet loss rate, bandwidth utilization rate and node load data of each index in the multidimensional performance index data) in the index change matrix.

[0049] The information entropy of each unit indicator is calculated by using the number of segments (i.e., the number of fixed time windows) and probability values ​​of the multidimensional performance index data divided by the fixed time window.

[0050] The information entropy of each unit indicator is filtered to obtain the minimum information entropy. The minimum information entropy is calculated using the inverse information entropy principle to obtain the initial weight coefficient. The initial weight coefficients of each row and column of the indicator change matrix are traversed to form a set of initial weight coefficients.

[0051] It should be noted that the index change matrix is ​​normalized, and the normalized values ​​are considered as part of the probability distribution. For example, the normalized value of unit index j in time window i is... Then the probability can be calculated: That is, for each unit index j, first sum its normalized values ​​over all time windows, then sum the normalized values ​​within each time window i. Divide by the sum of all time period values ​​of the unit indicator to obtain the probability. The purpose of calculating the probability is to enable unit indicators of different scales (i.e., unit indicators refer to the various indicators in multidimensional performance indicator data, such as transmission delay, packet loss rate, bandwidth utilization and node load data) to be compared with a unified standard, thereby calculating information entropy.

[0052] Then calculate the information entropy of each unit indicator. ; where the coefficient Normalization constant (e.g.) (For the number of time windows), ensure that the value is within the specified range. Within the range; the reason for choosing the minimum information entropy is that the smaller the information entropy (indicating that the data is more discriminative), the higher the corresponding weight of the indicator (or indicator unit), and vice versa. This ensures that the node load correction score more reasonably reflects the importance of each performance indicator to the performance evaluation, thus obtaining a lower initial weight coefficient.

[0053] S23: Record the collection timestamp of the multidimensional performance index data, and calculate the time difference using the current timestamp and the recorded collection timestamp (i.e., the time difference reflects the newness of the multidimensional performance index data).

[0054] A time decay coefficient is preset; the time decay factor is calculated using the time difference and the preset time decay coefficient (i.e., the time decay factor reflects the contribution value of the new and old multidimensional performance index data, that is, the weight level, and the initial weight coefficient can be updated later).

[0055] The multidimensional performance index data for each time stamp is analyzed to determine the node load data. The mean and standard deviation of the node load data within each time period sample are calculated (i.e., the mean and standard deviation reflect the average fluctuation level and degree of fluctuation of the node load data in each time period sample).

[0056] The volatility coefficient is calculated using the average value and standard deviation of the node load data within each time period sample; the corrected weight coefficient is calculated using the initial weight coefficient, the corresponding time decay factor, and the corresponding volatility coefficient to obtain a set of corrected weight coefficients.

[0057] It should be noted that calculating the time decay factor helps to consider the timeliness of the data, giving higher weight to newer data, so that the cloud computing system can pay more attention to the current node load status (that is, through the time decay factor, it can better adapt to changes in the network environment and performance fluctuations, and improve the real-time performance and reliability of path selection); step S23 also involves calculating the fluctuation coefficient, which can measure the fluctuation of node load, further affecting the accuracy of the node load correction score.

[0058] In step S24, the process mainly involves multiplying each unit index in the index change matrix with the set of correction weight coefficients one by one to obtain the weighted value of the multidimensional performance index data of each node at each time point; summing all the weighted values ​​to obtain the node load correction score of each node.

[0059] S24: The index change matrix is ​​normalized again using a hybrid normalization method. Each unit index in the index change matrix after the second normalization is multiplied one by one with the set of corrected weight coefficients to obtain the weighted value of the multidimensional performance index data at each time point of each node.

[0060] By summing the weighted values ​​of all multidimensional performance metrics at each time point, a node load correction score for each node is obtained (which facilitates a unified assessment of the relative performance of transmission paths or nodes). The calculation formula is as follows:

[0061] ;

[0062] In the formula, Represented as the first The initial baseline score (i.e., the benchmark score) of each node is determined by the index change matrix normalization and information entropy algorithm.

[0063] Represented as the first The The multi-dimensional performance metrics data (i.e., transmission delay, packet loss rate, bandwidth utilization, and node load data) are calculated based on the difference between the data acquisition time and the current time. Calculate the time decay factor;

[0064] Represented as the first The corrected weighting coefficients for the multidimensional performance index data take into account factors such as initial weights, time decay, and load fluctuations.

[0065] Represented as the first The load fluctuation coefficient of a multidimensional performance index data reflects the volatility of the index data; the greater the volatility, the lower the value.

[0066] Represented as the first The deviation between the ideal target value and the current actual measured value of a multidimensional performance index;

[0067] This represents the total number of multidimensional performance index data, used to normalize the contribution of each index;

[0068] It should be noted that re-normalization helps to handle biases in the data and ensures that all indicators are evaluated under a unified standard. The weighted value calculation, by multiplying the corrected weights by the normalized value of each indicator, more accurately reflects the overall performance of the nodes. It comprehensively considers factors such as time decay factors and volatility, thereby accurately scoring each node. This allows each node to fully reflect its performance fluctuations over different time periods during evaluation, further improving the scientific nature and adaptability of path selection.

[0069] S25: Use the TOPSIS algorithm to collect multi-dimensional performance index data from nodes of each historical transmission path to construct an ideal solution. Use the ideal solution and node load correction score to filter the first performance transmission path and backup path set. Construct an initial path selection strategy using the first performance transmission path and backup path set. Collect the original test data stream to reconstruct the initial path selection strategy to obtain the final path selection strategy.

[0070] It should be noted that the TOPSIS algorithm constructs an ideal solution and combines it with multi-dimensional performance index data and node load correction scores from the nodes of each historical transmission path to select the optimal transmission path. The implementation process of step S25 above can more clearly identify which paths have the best performance through the comparison of ideal solutions. The final path selection strategy is based on the adjustment of the initial path selection strategy and combined with actual test data to ensure the accuracy of transmission path selection.

[0071] Meanwhile, the TOPSIS algorithm also helps to compare the overall performance of various transmission paths and select the best path; by reconstructing the path selection strategy and combining actual traffic data and network conditions, it ensures the adaptability and real-time performance of the path selection strategy; through the optimization of the TOPSIS algorithm and the path selection strategy, the transmission efficiency of the network can be improved and the system stability can be ensured.

[0072] Specifically, such as Figure 5 As shown, in step S25, the TOPSIS algorithm is used to collect multi-dimensional performance index data from nodes of each historical transmission path to construct an ideal solution. The ideal solution and node load correction scores are then used to filter the first performance transmission path and backup path sets. An initial path selection strategy is constructed using the first performance transmission path and backup path sets. The original test data stream is collected to reconstruct the initial path selection strategy, resulting in the final path selection strategy. The specific operation steps are as follows:

[0073] In steps S251-S252, the process mainly describes how to use the TOPSIS algorithm to collect multi-dimensional performance index data of nodes in each historical transmission path to construct an ideal solution; calculate the proximity of each transmission path based on the ideal solution and the node load correction score; and select the first performance transmission path and the set of backup paths from the transmission paths based on the maximum proximity sieve and a preset dynamic threshold.

[0074] S251: Use the TOPSIS algorithm to collect multi-dimensional performance index data from nodes of each historical transmission path, and construct an ideal solution based on the historical multi-dimensional performance index data;

[0075] The ideal solution includes a positive ideal solution (i.e., the positive ideal solution may be the best node load correction score in the long-term monitoring of each indicator in the multidimensional performance index data) and a negative ideal solution (i.e., the negative ideal solution is the worst set of indicators in the multidimensional performance index data in theory).

[0076] The node load correction score of each node in the transmission path is used to calculate the Euclidean distance with the ideal solution to obtain the positive and negative distances.

[0077] The proximity of each transmission path is calculated using the positive and negative distances (i.e., the proximity represents the closest ideal solution; the higher the proximity, the better the path performance).

[0078] It should be noted that the TOPSIS algorithm evaluates the performance of each transmission path and selects the path closest to the ideal solution. The ideal solution includes a positive ideal solution (the best node load correction score, which represents the best performance among all historical multidimensional performance metrics of nodes across all transmission paths, and the highest performance metric value achieved by a node at any given time across all nodes) and a negative ideal solution (the worst set of metrics, which represents the worst performance among all historical multidimensional performance metrics of nodes across all transmission paths). The proximity of each path is calculated, with paths having higher proximity exhibiting better performance. By analyzing the multidimensional performance metrics of nodes across all historical transmission paths, the ideal solution is established and proximity is calculated to select the path with the best performance. This ensures the selection of the most stable and high-performance path, improving network transmission efficiency and reliability.

[0079] The TOPSIS algorithm is a common multidimensional data analysis method that can effectively integrate multiple performance indicators to help identify the optimal path. By setting positive and negative ideal solutions, it can ensure that the path selection takes into account various aspects of performance optimization.

[0080] S252: Sort each transmission path in descending order according to proximity to obtain a transmission path sequence list, and take the transmission path with the highest proximity as the first performance transmission path (that is, the transmission path with the best performance, which is also the main transmission path).

[0081] A preset dynamic threshold (i.e., the dynamic threshold is calculated based on collected historical network performance data, including network topology complexity (i.e., the complexity of the transmission path) and historical switching success rate (i.e., the success rate of switching transmission paths when transmission is not smooth) is used to determine whether the proximity of each of the other transmission paths in the transmission path sequence list, excluding the first performance transmission path, is greater than the dynamic threshold.

[0082] If so, the remaining transmission paths are determined as backup transmission paths (that is, if the proximity of the remaining transmission paths is greater than the dynamic threshold, it means that the remaining transmission paths have sufficient performance reserves and can be used as backup paths).

[0083] The remaining transmission paths that meet the proximity requirement greater than the dynamic threshold are grouped together to obtain a backup path set (when the primary transmission path (i.e., the first performance transmission path) fails, the backup transmission paths in the backup path set are used).

[0084] It should be noted that the paths are sorted according to proximity to select the best primary transmission path; a preset dynamic threshold is used to determine the availability of backup paths based on historical network performance data.

[0085] Ensure that both the primary and backup paths have sufficient performance reserves and can quickly switch to the backup path when the primary path fails; the setting of dynamic thresholds enables the network to make intelligent judgments based on the real-time changes in the network environment.

[0086] The introduction of dynamic thresholds can effectively address the complexity of network topology and improve network handover success rate, thereby enhancing network adaptability; the selection of backup paths ensures rapid recovery in case of failure, improving the network's fault tolerance.

[0087] In steps S253-S254, the main steps are to construct a topology network path graph using the first performance transmission path and the set of backup paths; to screen the shared nodes and shared links between each transmission path, and to calculate the overlap of the shared nodes and shared links; to use the overlap ratio to calculate the overlap of the transmission paths in the overlapping part and update the topology network path graph to obtain a path dependency graph; and to assign path priorities to the first performance transmission path and the set of backup paths in the path dependency graph.

[0088] An initial path selection strategy is generated based on the path priority of the first performance transmission path and the set of backup paths.

[0089] S253: Construct a topology network path diagram based on the first performance transmission path and the set of backup paths according to the cloud computing network;

[0090] The network path graph is evaluated to identify shared nodes and shared links between each transmission path (i.e., when all transmission paths are used to construct the topology graph, there may be overlapping transmission paths, partial overlap of paths, or overlap of nodes).

[0091] Calculate the overlap between the shared node and the shared link (i.e., the overlap can be calculated using Euclidean distance).

[0092] A preset overlap threshold is set; it is then determined whether the overlap between the shared node and the shared link is greater than the overlap threshold.

[0093] If so, the shared node and the shared link are determined to be overlapping, and the overlap ratio of the corresponding transmission path is calculated for the overlapping shared node and the shared link (that is, the overlap ratio of the overlapping part of the shared node and the shared link to the corresponding total transmission path).

[0094] Set a preset overlap ratio threshold; determine whether the overlap ratio is greater than the preset overlap ratio threshold;

[0095] If so, remove the overlapping shared nodes and shared links (i.e., remove the overlapping parts because they have too much redundancy), and update the topology network path graph to obtain the path dependency graph.

[0096] It should be noted that the process involves constructing a network path graph, evaluating the overlap of shared nodes and links between paths, removing redundant parts based on a set overlap threshold, optimizing path selection, evaluating the shared parts of different transmission paths to avoid unnecessary overlap and reduce path redundancy, and optimizing network paths and improving transmission efficiency by removing overlapping parts.

[0097] When multiple paths overlap, removing redundant parts can avoid unnecessary resource waste and performance degradation; the optimized path selection is more flexible and avoids network paralysis caused by a single path failure.

[0098] S254: The first performance transmission path in the path dependency graph is taken as the first path priority (that is, the highest path priority). The set of backup paths (that is, the backup transmission paths at this time with overlapping parts removed) are sorted according to their proximity and given the second path priority... the nth path priority (that is, to ensure that the selected backup path and the main path have low overlap in physical links or logical nodes).

[0099] The node load correction scores of each node in the first performance transmission path are summed to obtain the path score (that is, the average score of the entire path).

[0100] Collect historical abnormal data, and set a preset abnormal threshold based on the historical abnormal data (i.e., set an abnormal threshold based on historical abnormal data such as sudden increase in latency and sharp increase in packet loss rate).

[0101] Determine whether the path score is greater than an anomaly threshold;

[0102] If so, then switch to an alternative transmission path based on the path priority of proximity (that is, switch the first performance transmission path with the first path priority to the second path priority, ... the nth path priority, indicating that the first performance transmission path may have encountered an anomaly during transmission).

[0103] An initial path selection strategy is generated by switching transmission paths based on the path priority of proximity.

[0104] It should be noted that the path dependency graph is optimized by sorting the paths according to their priority. This priority sorting ensures low overlap between backup paths and primary paths, thus optimizing the path selection strategy. During path selection, the physical or logical overlap between primary and backup paths is kept as low as possible to avoid redundant transmission bottlenecks. Optimizing path selection enables the network to quickly switch to the optimal path in the event of a failure.

[0105] Ensure that path selection has high flexibility and fault recovery capability in the case of multi-path redundancy; through reasonable path dependency, network conflicts can be reduced and transmission stability and efficiency can be improved.

[0106] Meanwhile, the newly collected historical anomaly data mainly reflects the performance of nodes on each transmission path in the network across multiple dimensions (such as latency, packet loss rate, throughput, etc.) over a period of time. It includes the performance of each node at different points in time and also provides the time-series performance of each node. That is, the multidimensional performance index data of each node changes over time and is mainly collected as a threshold. The multidimensional performance index data of each node collected in step S251 mainly reflects the performance data of each node on different paths in the network, including the values ​​of multiple indicators. Each node may have different performance at different times and under different conditions. These performance indicators can be recorded through historical anomaly data. It is also a sample at a specific point in time, containing the various performance indicators of that node at a specific moment.

[0107] S255: Collect the original test data stream and divide it into data stream replicas. Analyze the multi-dimensional performance index data using the data stream replicas to obtain a new path load score. Adjust the path priority of each transmission path in the initial path selection strategy using the new path load score to obtain a second path selection strategy. Instantaneous sampling of each node is performed using the new first performance transmission path in the second path selection strategy, and a preset dynamic threshold is determined to filter the final backup transmission path. The final path selection strategy is formed by combining the new first performance transmission path and the final backup transmission path.

[0108] It should be noted that the original test data stream is analyzed and path priorities are adjusted; the path selection strategy is dynamically adjusted by evaluating the new path load score of each path to ensure that the optimal path can be used under any circumstances.

[0109] By analyzing data stream replicas and node load correction scores, the path selection strategy is adjusted in real time to improve transmission stability and ensure that the final selected path maintains efficient transmission under different conditions.

[0110] Real-time analysis of multidimensional performance data ensures the flexibility of path selection strategies to adapt to constantly changing network environments; dynamic adjustment of strategies can minimize performance degradation caused by network failures.

[0111] While the above solutions consider the selection and updating of backup paths, they may not fully address issues such as resource reuse and path independence. If backup paths share a large number of nodes or links with the primary path, their performance and stability may be significantly compromised. When the primary path fails or its performance degrades, the system needs to switch to the backup path as quickly as possible. Significant delays during path selection may affect the network's timely response, leading to a decline in user experience, especially in scenarios with high real-time requirements.

[0112] Therefore, to optimize the path selection strategy and make it more accurate, dynamic, and robust, the following steps ensure broader test coverage across multiple paths by collecting and replicating raw test data streams, more realistically reflecting the network environment and multi-path performance; utilizing methods such as sliding window dynamic calculation, the path selection strategy can be flexibly adjusted according to changes in network conditions; this mechanism can handle various dynamic situations such as network traffic changes and device failures, ensuring the network always operates in optimal condition; and dynamically selecting and switching to backup paths ensures that the system can quickly respond and switch to backup paths when the primary path encounters problems, maintaining network stability and avoiding prolonged service interruptions; with the adjustment of path priorities and load correction, the use of each path can be optimized, resulting in more efficient allocation of network resources, reducing resource waste and congestion, and improving the overall network transmission efficiency; and through real-time sampling and dynamic threshold setting of backup paths, the S255 extension ensures that backup paths can quickly take over when network failures occur, ensuring uninterrupted user traffic, more stable and reliable network services, and optimizing backup transmission paths;

[0113] Specifically, such as Figure 6 As shown, in step S255, the original test data stream is collected and divided into data stream replicas. The data stream replicas are used to analyze the multi-dimensional performance index data to obtain a new path load score. The path priority of each transmission path is adjusted using the new path load score to obtain a second path selection strategy. The new first performance transmission path in the second path selection strategy is used to perform instantaneous sampling of each node and determine a preset dynamic threshold to screen the final backup transmission path. The new first performance transmission path and the final backup transmission path constitute the final path selection strategy. The specific operation steps are as follows:

[0114] Steps S2551-S2553 mainly describe the process of collecting the original test data stream and replicating it using a hierarchical traffic mirroring method to obtain multiple data stream replicas; adding a sequence identifier to each data stream replica; pre-setting collection and monitoring indicators to collect data on the first performance transmission path and the backup transmission path in the initial path selection strategy; comparing the collected data with the sequence identifier of each data stream replica to obtain a high-precision response dataset; and updating the multi-dimensional performance indicator data based on the high-precision response dataset to obtain a multi-dimensional performance indicator data set.

[0115] The node load correction score is recalculated on the multidimensional performance index data set; the path stability coefficient of the transmission path is calculated based on the recalculated node load correction score; and a new path load score of the transmission path is obtained by using the recalculated node load correction score and the path stability coefficient of the transmission path.

[0116] S2551: Collect raw test data stream (i.e., the same data used for transmitting data to simulate real network transmission and to test the initial path selection strategy).

[0117] The original test data stream is copied using a hierarchical traffic mirroring method to obtain multiple data stream replicas, the number of which corresponds to the number of transmission paths;

[0118] Add a sequence identifier to each data stream replica (i.e., by appending a multi-dimensional performance metric (i.e., transmission delay, packet loss rate, bandwidth utilization, and node load data) to the packet header);

[0119] Add a timestamp to each data stream replica to mark the start time of the test initial path selection strategy;

[0120] It should be noted that collecting raw test data streams to simulate real network transmission conditions provides reliable test data for path selection strategies; replicating data streams using a hierarchical traffic mirroring method, and ensuring the number of replicas matches the number of transmission paths, achieves higher test accuracy; the processing in step S2551 ensures that the simulated transmission environment is consistent with the actual network, and testing multiple paths through replicas improves the comprehensiveness of the test, such as... Figure 7 As shown, this provides a data source for the subsequent collection of multidimensional performance indicators;

[0121] S2552: Based on the timestamp synchronization method, each data stream replica is synchronized and distributed to the first performance transmission path and the backup transmission path in the initial path selection strategy; and the collection and monitoring indicators (i.e., end-to-end latency, packet loss distribution, and node load peak) are set.

[0122] Data is collected for each data stream replica on each transmission path according to the collected monitoring indicators (i.e., end-to-end latency, packet loss distribution, and node load peak).

[0123] The data collected on each transmission path is compared with the sequence identifier of each data stream replica (i.e., matching can be done by calculating Euclidean distance) to obtain real-time data (i.e., data that closely reflects actual transmission delay, packet loss rate, bandwidth utilization, and node load). The real-time data on each transmission path are then aggregated to obtain a high-precision response dataset.

[0124] By fusing the high-precision response dataset with the multidimensional performance index data, the multidimensional performance index data is updated to obtain a multidimensional performance index data set.

[0125] It should be noted that the execution process of step S2553 above uses timestamp synchronization to ensure that each data stream copy is sent at the same start time in all transmission paths, and sets monitoring indicators to collect real-time network performance data to ensure synchronized packet sending, which can provide consistency for subsequent comparison of the performance of different paths; and performs real-time monitoring of key performance indicators (latency, packet loss, load) to obtain detailed performance datasets.

[0126] S2553: Recalculate the node load correction score for each transmission path based on the multidimensional performance index dataset (i.e., the calculation steps for the node load correction score have been explained in S24 and will not be repeated here).

[0127] The dynamic rate of change of a multidimensional performance index dataset is recalculated using a sliding window.

[0128] The switching frequency of the index within each sliding window is determined using the recalculated dynamic change rate for the multidimensional performance index data set (that is, with each sliding window as a period, the switching frequency of each index as a single transmission is determined for each transmission path based on the recalculated dynamic change rate).

[0129] The dynamic fluctuation amplitude between sliding windows is calculated using the index switching frequency within each sliding window to analyze the multidimensional performance index dataset.

[0130] The path stability coefficient is obtained by weighting the indicator switching frequency with the dynamic fluctuation amplitude.

[0131] The recalculated node load correction score is corrected using the path stability coefficient, and a new path load score is obtained by calculating the entire transmission path (i.e., the score of the nodes in the transmission path after correction, and then the comprehensive new path load score of the entire transmission path is calculated using the corrected scores of all nodes, which is the new path score in step S253, the score of the new transmission path).

[0132] It should be noted that by using a multi-dimensional performance index dataset and dynamic change rate, the new path load score for each path is recalculated to ensure more accurate path evaluation; real-time data is integrated to ensure that the node load correction score can reflect the actual changes in network performance; and the stability of the path is calculated using a sliding window and dynamic change rate to optimize path priority selection.

[0133] S2554: Adjust the path priority of each transmission path using the new path load score (note that it is different from the path score in step S253 above) to generate a new path priority strategy (i.e., re-evaluate the first path priority path and the backup transmission path for each transmission path based on the new path load score).

[0134] The path dependency graph is reconstructed using the new path priority strategy to obtain a new path dependency graph;

[0135] The initial path selection strategy is updated using the comprehensive score, stability coefficient, and new path dependency graph to generate a second path selection strategy;

[0136] It should be noted that the path priority is adjusted by using node load correction scoring and the path dependency graph is reconstructed to generate an updated path selection strategy; the path priority is optimized to ensure that the best-performing path is selected; and by reconstructing the path dependency graph, possible path conflicts and redundancies are eliminated, thereby improving overall transmission efficiency.

[0137] In step S2555, the main steps are to deploy distributed probes on each node of the new first performance transmission path in the second path selection strategy to perform instantaneous sampling at a preset sampling frequency; and to judge the instantaneous sampling based on a preset dynamic threshold to select the final backup transmission path.

[0138] S2555: Deploy distributed probes for each node in the new first performance transmission path (i.e., the first path priority path, where the first performance transmission path is the transmission path updated by the new path priority strategy) in the second path selection strategy.

[0139] The sampling frequency is set for each node of the new first performance transmission path, and the multi-dimensional performance index data set is sampled instantaneously at each node (i.e., timestamp, latency value, packet loss ratio, CPU / memory / network load).

[0140] A preset dynamic threshold is set; it is determined whether the instantaneous sampling of any unit indicator of each node (i.e., the transmission delay, packet loss rate, bandwidth utilization rate and node load data of each indicator of multidimensional performance data) is greater than the dynamic threshold.

[0141] If so, then a new backup transmission path is selected from the set of backup paths in the second path selection strategy according to the new path priority strategy (that is, the backup transmission path with the highest priority is selected from the set of backup paths according to the new path load score and switched to take over the first performance transmission path as the second path priority).

[0142] The reconstructed path dependency graph is used to re-detect the shared nodes of the new backup transmission path and calculate the overlap ratio. The final backup transmission path is determined by the overlap ratio (i.e., see step S253 for shared node screening and overlapping node calculation, which will not be repeated here; using the pre-constructed path dependency graph, the candidate backup paths are topologically conflict-detected; paths that have a lot of shared links or key nodes that overlap with the existing main transmission path and have insufficient redundancy are eliminated to ensure that the final selected backup path has low risk and high independence).

[0143] It should be noted that the newly selected path is monitored in real time through instantaneous sampling, and backup paths are selected based on dynamic thresholds; path performance is monitored in real time to quickly respond to emergencies and ensure the stability of the first-performing path; backup paths are selected based on real-time sampling and dynamic thresholds to ensure that the network can quickly switch to the optimal path in dynamic changes.

[0144] S2556: The new first performance transmission path and the final backup transmission path are used to reconstruct the final path selection strategy.

[0145] It should be noted that by combining the newly selected primary transmission path and the backup transmission path, a final path selection strategy is constructed to ensure that the path selection is both efficient and stable; the final path selection strategy can guarantee robustness in complex network environments; and ensure that even if the primary transmission path has problems, the backup path can take over in time to maintain network performance.

[0146] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; those skilled in the art can modify the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A system for evaluating and assessing the performance indicators of cloud computing network data transmission, characterized in that, This includes a data center sampling and analysis module and a node load analysis and correction processing module: The data center sampling and analysis module is used to collect transmission data from nodes along various transmission paths in the cloud computing network; preprocess the transmission data to obtain multi-dimensional performance index data; the multi-dimensional performance index data includes transmission delay, packet loss rate, bandwidth utilization, and node load data; The node load analysis and correction processing module is used to divide the multidimensional performance index data into time period samples and construct an index change matrix for each time period sample. The multidimensional performance index data is recorded and collected with timestamps. Node load data is analyzed for each timestamp of the multidimensional performance index data. A time decay coefficient is introduced using the node load data within each time period sample to calculate a set of corrected weight coefficients. The set of corrected weight coefficients is used to correct the index change matrix to obtain a node load correction score. The node load correction scores of nodes along each transmission path are used to select the preferred transmission path, thus forming a path selection strategy. The multidimensional performance index data is processed into a sliding window to obtain samples for multiple time periods; Construct an indicator change matrix using the samples from the multiple time periods; The index change matrix is ​​normalized, and the information entropy of each unit index is calculated. The minimum information entropy is selected, and the initial weight coefficient of each unit index in each row and column of the index change matrix is ​​calculated using the inverse information entropy principle to obtain the initial weight coefficient set. The multidimensional performance index data is recorded and collected with timestamps. For each timestamp, the multidimensional performance index data is analyzed using node load data, and the fluctuation coefficient of the node is calculated. By using the fluctuation coefficient and the initial weight coefficient set, and introducing the time decay coefficient, a modified weight coefficient set is obtained; The weighted values ​​of the multidimensional performance index data at each time point of each node are obtained by multiplying each unit index in the index change matrix with the set of corrected weight coefficients one by one. Summing all weighted values ​​yields a node load correction score for each node. The TOPSIS algorithm is used to collect multi-dimensional performance index data from nodes along historical transmission paths to construct an ideal solution. This ideal solution, along with the node load correction score, is used to select a set of first-performance transmission paths and backup paths. An initial path selection strategy is constructed using the first-performance transmission paths and backup paths. The original test data stream is collected to reconstruct the initial path selection strategy, resulting in the final path selection strategy.

2. A method for evaluating and processing performance indicators of cloud computing network data transmission, characterized in that, It employs the cloud computing network data transmission performance index identification and evaluation processing system as described in claim 1 to perform the evaluation process, including the following steps: The transmission data of each node in the cloud computing network is collected; the transmission data is preprocessed to obtain multi-dimensional performance index data; the multi-dimensional performance index data includes transmission delay, packet loss rate, bandwidth utilization rate and node load data. The multidimensional performance index data is divided into time period samples, and an index change matrix is ​​constructed for each time period sample; The multidimensional performance index data is recorded and collected with timestamps. Node load data is analyzed for each timestamp of the multidimensional performance index data. A time decay coefficient is introduced using the node load data within each time period sample to calculate a set of corrected weight coefficients. The set of corrected weight coefficients is used to correct the index change matrix to obtain a node load correction score. The node load correction scores of nodes along each transmission path are used to select the preferred transmission path, thus forming a path selection strategy. The multidimensional performance index data is divided into time period samples, and an index change matrix is ​​constructed for each time period sample; The multidimensional performance index data is recorded and collected using timestamps. For each timestamp, the node load data of the multidimensional performance index data is analyzed. A time decay coefficient is introduced using the node load data within each time period sample to calculate and obtain a set of corrected weighting coefficients. The specific operation steps are as follows: The multidimensional performance index data is processed by a sliding window to obtain multiple time period samples; an index change matrix is ​​constructed using the multiple time period samples; the index change matrix is ​​normalized, and the information entropy of each unit index is calculated; the minimum information entropy is selected and the initial weight coefficient of each unit index in each row and column of the index change matrix is ​​calculated using the inverse information entropy principle to obtain the initial weight coefficient set. The multidimensional performance index data is recorded and collected with timestamps. For each timestamp, the multidimensional performance index data is analyzed using node load data, and the fluctuation coefficient of the node is calculated. By using the fluctuation coefficient and the initial weight coefficient set, and introducing the time decay coefficient, a set of corrected weight coefficients is obtained; The weighted values ​​of the multidimensional performance index data at each time point of each node are obtained by multiplying each unit index in the index change matrix with the set of corrected weight coefficients one by one. Sum all weighted values ​​to obtain the node load correction score for each node; The TOPSIS algorithm is used to collect multi-dimensional performance index data from nodes of each historical transmission path to construct an ideal solution. The ideal solution and node load correction score are used to screen the first performance transmission path and the set of backup paths. An initial path selection strategy is constructed using the first performance transmission path and the set of backup paths. The original test data stream is collected to reconstruct the initial path selection strategy to obtain the final path selection strategy.

3. The method for evaluating and processing cloud computing network data transmission performance indicators according to claim 2, characterized in that, The TOPSIS algorithm is used to collect multi-dimensional performance index data from nodes along various historical transmission paths to construct an ideal solution. This ideal solution, combined with node load correction scores, is then used to select the first-performing transmission path and a set of backup paths. The specific steps are as follows: The TOPSIS algorithm is used to collect multidimensional performance index data from nodes of each historical transmission path to construct an ideal solution; the proximity of each transmission path is calculated based on the ideal solution and the node load correction score. The first performance transmission path and the set of backup paths are selected from the transmission paths based on the maximum proximity sieve and the preset dynamic threshold.

4. The method for evaluating and processing cloud computing network data transmission performance indicators according to claim 3, characterized in that, An initial path selection strategy is constructed by combining the first performance transmission path with the set of backup paths. The specific steps are as follows: A topology network path graph is constructed using the first performance transmission path and the set of backup paths; shared nodes and shared links between each transmission path are selected, and the overlap of shared nodes and shared links is calculated; the overlap ratio of the overlapping transmission paths is calculated using the overlap ratio to update the topology network path graph, resulting in a path dependency graph. The path priority is assigned to the first performance transmission path and the set of backup paths in the path dependency graph; An initial path selection strategy is generated based on the path priority determined by the set of first performance transmission paths and backup paths.

5. The method for evaluating and processing cloud computing network data transmission performance indicators according to claim 4, characterized in that, The initial path selection strategy is reconstructed by collecting the original test data stream to obtain the final path selection strategy. The specific operation steps are as follows: The original test data stream is collected and divided into data stream replicas. The data stream replicas are used to analyze the multidimensional performance index data to obtain a new path load score. The new path load score is used to adjust the path priority of each transmission path in the initial path selection strategy to obtain a second path selection strategy. The new first performance transmission path in the second path selection strategy is used to perform instantaneous sampling of each node and determine a preset dynamic threshold to screen the final backup transmission path. The new first performance transmission path and the final backup transmission path constitute the final path selection strategy.

6. The method for evaluating and processing cloud computing network data transmission performance indicators according to claim 5, characterized in that, The original test data stream is collected and divided into data stream replicas. The data stream replicas are then used to analyze the multi-dimensional performance index data to obtain the new path load score. The specific operation steps are as follows: The original test data stream is collected and replicated using a hierarchical traffic mirroring method to obtain multiple data stream replicas; a sequence identifier is added to each data stream replica; monitoring indicators are pre-set to collect data on the first performance transmission path and the backup transmission path in the initial path selection strategy; the collected data is compared with the sequence identifier of each data stream replica to obtain a high-precision response dataset. The multidimensional performance index data is updated based on the high-precision response dataset to obtain a multidimensional performance index data set. The node load correction score is recalculated on the multidimensional performance index dataset; The path stability coefficient of the transmission path is calculated based on the recalculated node load correction score; the new path load score of the transmission path is obtained by using the recalculated node load correction score and the path stability coefficient of the transmission path.

7. The method for evaluating and processing cloud computing network data transmission performance indicators according to claim 6, characterized in that, The second path selection strategy is obtained by adjusting the path priority of each transmission path using the new path load score. The specific operation steps are as follows: The path priority of each transmission path is adjusted using the new path load score to generate a new path priority strategy; the path dependency graph is reconstructed using the new path priority strategy to obtain a new path dependency graph; the initial path selection strategy is updated using the comprehensive score, stability coefficient, and the new path dependency graph to generate a second path selection strategy.

8. The method for evaluating and processing cloud computing network data transmission performance indicators according to claim 7, characterized in that, Instantaneous sampling of each node is performed using the new first performance transmission path in the second path selection strategy, and a preset dynamic threshold is determined to filter the final backup transmission path. The final path selection strategy is constructed by combining the new first performance transmission path and the final backup transmission path. The specific operation steps are as follows: Distributed probes are deployed to each node in the new first performance transmission path in the second path selection strategy to perform instantaneous sampling at a pre-set sampling frequency; the instantaneous sampling is judged according to a preset dynamic threshold to select the final backup transmission path; The new first performance transmission path and the final backup transmission path are used to reconstruct the final path selection strategy.

Images