Cross-cloud resource migration method and system

Abstract

The invention relates to the technical field of cloud computing resource migration, in particular to a cross-cloud resource migration method and system, and the method comprises the following steps: carrying out the proportional distribution of the processing capability of a virtual machine instance based on the processing capability of the virtual machine instance, the block storage read-write rate and the network delay time, carrying out weighted average on the block storage read-write rate and the network delay time, integrating the block storage read-write rate and the network delay time into a two-dimensional matrix, sorting and removing low-threshold nodes, and generating a preliminary resource optimization data set; according to the method, resource optimization accuracy is improved through proportional allocation of virtual machine processing capacity, weighted average block storage read-write rate and network delay and performance index matrix processing, performance weight difference and path parameter optimization performance dependency expression are combined, node transmission efficiency and priority are dynamically adjusted, transmission path elasticity is enhanced, and resource optimization efficiency is improved. The resource state of the target end is monitored and adjusted in real time, migration stability is ensured, the resource integration precision and service controllability of the multi-cloud environment are improved, and complex application requirements are met.

Description

Technical Field

[0001] The present invention relates to the technical field of cloud computing resource migration, and in particular to a cross-cloud resource migration method and system. Background Art

[0002] The field of cloud computing resource migration technology mainly studies how to achieve efficient migration and management of resources (including data, applications, computing tasks, etc.) between different cloud computing environments (such as public clouds, private clouds, and hybrid clouds). This technology field aims to solve the problems of resource scheduling, data consistency, system compatibility, etc. faced by enterprises in multi-cloud or hybrid cloud environments. By using automation tools, containerization technology, and cross-platform compatibility technology, the resource migration process is faster, safer, and more controllable. This field has a wide range of application scenarios, including enterprise business expansion, disaster recovery switching, load balancing, and cost optimization, helping enterprises to achieve efficient integration and utilization of resources in a multi-cloud strategy.

[0003] Among them, the cross-cloud resource migration method is a method and system for efficiently migrating resources between different cloud environments. By designing tools and strategies that adapt to multi-cloud environments, it can achieve flexible resource scheduling, seamless migration, and secure data transmission, avoiding business interruptions caused by differences in cloud service providers or compatibility issues. Its purpose is to help enterprises realize dynamic allocation and optimization of resources in multi-cloud or hybrid cloud environments, improve business flexibility and scalability, and support digital transformation and continuous innovation and development of business.

[0004] In the process of cloud resource migration, the existing technologies do not adequately handle the compatibility issues of different cloud environments, lack deep integration of multiple performance indicators such as virtual machine instances, storage read and write rates, and network latency, resulting in insufficient resource scheduling accuracy and a preference process that relies heavily on empirical judgment. The description of resource performance dependency paths is too simplistic, making it difficult to fully reflect the complex dependencies between multiple nodes, resulting in resource conflicts or path failures during the migration process. When adjusting the resource migration path, the existing technologies lack a dynamic priority adjustment mechanism for high-load nodes, which easily causes increased data transmission delays and performance degradation. The monitoring and expansion of resource status also lack real-time performance, often resulting in the inability to discover and correct performance offset problems in a timely manner, affecting the load balancing and stable operation of nodes after the migration is completed. For example, during cross-cloud migration, the lagging monitoring of resource status causes some nodes in the migration path to be overloaded, ultimately leading to low migration efficiency and business interruption, limiting the application adaptation and migration efficiency of existing technologies in multi-cloud or hybrid cloud environments, and their dynamic response capabilities to business needs are relatively weak. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a cross-cloud resource migration method and system.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: a cross-cloud resource migration method, comprising the following steps: S1: Based on the processing capacity of the virtual machine instances, the block storage read / write rate, and the network delay time, the processing capacity of the virtual machine instances is proportionally allocated, the block storage read / write rate and the network delay time are weighted averaged, integrated into a two-dimensional matrix, sorted and removed low-threshold nodes, and a preliminary resource optimization data set is generated; S2: Based on the preliminary resource optimization data set, extract the performance weight difference between the virtual machine and the private subnet, accumulate the distributed file dependency path parameters to the node path, generate the resource performance dependency weight value, and fill it into the mapping relationship graph path set, perform optimal connection screening on the key path, and generate the resource mapping optimization path set; S3: Based on the resource mapping optimization path set, the transmission delay of the virtual machine and the bandwidth occupation of the load balancer are analyzed, the memory idle rate is matched, the transmission efficiency distribution value is recursively adjusted, the transmission priority is reallocated to the nodes exceeding the critical point, the transmission priority is iteratively adjusted, and a resource priority migration sequence table is generated; S4: Based on the resource priority migration sequence table, the target virtual machine operating status and network delay are updated in real time, the memory usage and read and write rates are compared, resources are expanded for nodes whose difference values ​​exceed the range, the target node performance offset value is corrected to within the standard range, and a migration completion node performance status table is generated. The node operating load is monitored and adjusted to generate a migration completion monitoring result.

[0007] As a further solution of the present invention, the steps of obtaining the preliminary resource optimization data set are specifically: S111: Proportionally allocate the processing capacity of the virtual machine instance based on the processing capacity of the virtual machine instance, the block storage read and write rate, and the network delay time, and generate a processing capacity weight value of the virtual machine instance by comparing the load conditions and required resources of the differentiated virtual machine instances; S112: Based on the processing capacity weight value of the virtual machine instance, a weighted average calculation is performed on the block storage read / write rate and the network delay time, using the formula: ; Generate weighted block storage and network latency weight values; in, is the weighted average, is the weight coefficient of the block storage read and write rate, is the weight coefficient of network delay time, is the block storage read and write rate, is the network delay time; S113: Integrate the weighted block storage and network delay weight values ​​and the processing capacity weight into a two-dimensional matrix, sort according to the weighted results, screen qualified virtual machine instances, remove instances below the threshold, and generate a preliminary resource optimization data set.

[0008] As a further solution of the present invention, the step of obtaining the resource performance dependency weight value is specifically as follows: S211: Based on the preliminary resource optimization data set, extract the performance indicator data between the virtual machine and the private subnet, perform a ratio analysis on the network traffic of the virtual machine and the subnet traffic, analyze the difference between the traffic ratio and the throughput and delay parameters, and generate a performance weight difference value between the virtual machine and the private subnet; S212: Based on the performance weight difference between the virtual machine and the private subnet, call the path parameter set of the distributed file, identify the node path weight and bandwidth ratio in the path, superimpose the influence of the number of path dependencies, and use the formula: ; Calculate the node path dependency weight; in, represents the node path dependency weight, Indicates the path The path weight of each node, Indicates the path The bandwidth share of each node, represents the cumulative number of path dependencies, Represents the total number of nodes in the path, is the weight adjustment coefficient; S213: Based on the node path dependency weight, combined with the performance weight difference value between the virtual machine and the private subnet, the weight data of the two are normalized, the weight dependency of the overall resource performance is analyzed, and the resource performance dependency weight value is generated.

[0009] As a further solution of the present invention, the step of obtaining the resource mapping optimization path set is specifically: S221: extracting each performance parameter in the resource set based on the resource performance dependency weight value, determining the weight difference of each pair of resources by comparing the performance values ​​between the resources, and classifying and counting the weight differences to generate a resource weight mapping node set; S222: Based on the resource weight mapping node set, by screening the node and path weight combination, analyzing the path weight accumulation value one by one, calculating the dependency strength of the associated resource node combination, matching the calculation result with the weight value, and generating a key resource path set; S223: Based on the key resource path set, by comparing the dependency weight relationship of the nodes in the path, the path segment with the optimal weight value is screened and connected, the integrity of the connected path is verified and the structure is adjusted, the optimized path is added to the mapping set, and the resource mapping optimized path set is obtained.

[0010] As a further solution of the present invention, the step of obtaining the resource priority migration sequence table is specifically as follows: S311: based on the resource mapping optimization path set, extract the virtual machine transmission delay data, calculate the average delay value of each path, and perform coupling analysis with the subnet load balancer bandwidth occupancy value, evaluate the transmission delay characteristics, select the path exceeding the target threshold, record the bandwidth occupancy ratio, and generate the analysis results of transmission delay and bandwidth occupancy; S312: Based on the analysis results of the transmission delay and bandwidth occupancy, the relationship between the path transmission load and the memory occupancy is evaluated, and the transmission efficiency distribution value is recursively adjusted using the formula: ; generating a distribution adjustment result of path transmission efficiency; in, represents the transmission efficiency distribution value, For the The memory idle rate of the path, is the path bandwidth occupancy weight, is the path load value, is the adjustment factor, is the sum of the path transmission priority weights, Indicates the total number of paths; S313: Based on the distribution adjustment result of the path transmission efficiency, reallocate priorities to the path nodes whose transmission efficiency exceeds the critical point, recursively adjust the node weights item by item according to the priority, sort the nodes according to their criticality and load priority, determine the migration priority, and generate a resource priority migration sequence table.

[0011] As a further solution of the present invention, the step of obtaining the migration completed node performance status table is specifically as follows: S411: Based on the resource priority migration sequence table, monitor the running status of the target virtual machine in real time, collect the network delay value and CPU occupancy rate of the target virtual machine, record the change data in combination with the node resource index, normalize the network delay and CPU occupancy rate, and obtain the real-time performance status of the target virtual machine; S412: Based on the real-time performance status of the target virtual machine, compare the memory usage rate and the read / write rate, determine the change in the mean memory usage through the difference value, calculate the performance deviation in combination with the read / write rate fluctuation range, correct the difference value, and adjust the resource expansion parameters using the formula: ; Generate a performance deviation correction result of the target virtual machine; in, Indicates the performance deviation value, is the current memory usage, is the target memory usage, is the current read and write rate, is the target read and write rate; S413: Based on the performance deviation correction result of the target virtual machine, perform resource expansion to adjust the nodes that exceed the standard range, correct the real-time performance status of the expanded nodes, normalize and adjust the performance values ​​to the standard range, summarize the node performance correction status, and generate a migration completed node performance status table.

[0012] As a further solution of the present invention, the steps for obtaining the migration completion monitoring result are specifically as follows: S421: Based on the migration completed node performance status table, extract node performance parameter data, read and filter the operation indicators node by node, perform time series analysis on CPU occupancy, memory usage, and network traffic values, and generate a node performance status record table; S422: Based on the node performance status record table, analyze the node load occupancy, mark and classify the nodes that exceed the load threshold, transfer the high-load node tasks to the low-load nodes, evaluate the low-load node performance indicators, verify the adjusted node load parameters, and obtain the node load adjustment results; S423: Based on the node load adjustment result, monitor the adjusted nodes in real time, capture the operating load and performance parameter data, compare the load stability and performance item by item, record the dynamic information of the node after migration, summarize the monitoring data, and generate the migration completion monitoring result.

[0013] A cross-cloud resource migration system, the cross-cloud resource migration system is used to execute the above cross-cloud resource migration method, the system comprises: The node screening module calculates the weighted value of the block storage read / write rate and the network delay time based on the processing capacity of the virtual machine instance, the block storage read / write rate, and the network delay time, combines the three parameters in proportion for evaluation, and sorts the nodes, screening the nodes with scores greater than the set threshold to obtain a set of high-scoring nodes; The path dependency module calculates the weight of the distributed file dependency path node attributes based on the high-score node set, accumulates the file path length to the node weight value, selects the path based on the performance difference between the virtual machine and the subnet, and generates a path weight optimization result; The transmission priority module analyzes the transmission delay and bandwidth occupancy ratio of the node based on the path weight optimization result, re-analyzes the transmission efficiency distribution in combination with the node memory idle rate, adjusts the priority of the node whose occupancy rate exceeds the threshold, and generates an optimized transmission path sequence; The migration performance monitoring module monitors the target virtual machine's operating status and network delay in real time based on the optimized transmission path sequence, compares the node memory usage and block storage read and write rate difference, corrects performance offsets and records performance status, and generates migration completion monitoring results.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, by proportionally allocating virtual machine processing power, weighted average block storage read and write rate and network delay time, the performance indicators of complex resources are uniformly processed in a matrix manner, the accuracy and efficiency of resource optimization are improved, and the performance weight difference and the cumulative calculation of path parameters are combined to optimize the expression of performance dependence, effectively avoiding the parameter fuzziness problem in resource scheduling, and strengthening the visualization and operability of dependency. By dynamically adjusting the transmission efficiency distribution and priority redistribution of nodes, the elastic regulation ability of the transmission path is enhanced, the problem of high-load node resource transmission bottleneck is solved, and the real-time status monitoring and dynamic expansion adjustment of target end resources are carried out to control the node performance offset and operation load deviation within a reasonable range, ensuring the stability and data consistency of the resource migration process. This optimization process based on dynamic calculation and step-by-step recursive adjustment realizes flexible scheduling and efficient migration of resources in a complex multi-cloud environment, and significantly improves the accuracy of resource integration and the controllability of business operation. The overall solution has significant gains in performance optimization, resource matching and migration reliability compared with the existing technology, and can meet the dynamic needs of complex application scenarios in a multi-cloud environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic diagram of the workflow of the present invention; Figure 2 A flowchart of the preliminary resource optimization data set in the present invention; Figure 3 It is a flow chart of resource performance dependence weight value in the present invention; Figure 4 A flow chart of a resource mapping optimization path set in the present invention; Figure 5 It is a flow chart of the resource priority migration sequence table in the present invention; Figure 6 A flowchart of a node performance status table for completing migration in the present invention; Figure 7 The flowchart of the monitoring result of migration completion in the present invention. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0017] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0018] Example 1: Please refer to Figure 1 The present invention provides a technical solution: a cross-cloud resource migration method, comprising the following steps: S1: Based on the processing capacity of the virtual machine instances, the block storage read / write rate, and the network delay time, the processing capacity of the virtual machine instances is proportionally allocated, the block storage read / write rate and the network delay time are weighted averaged, integrated into a two-dimensional matrix, sorted and removed low-threshold nodes, and a preliminary resource optimization data set is generated; S2: Based on the preliminary resource optimization data set, the performance weight difference between the virtual machine and the private subnet is extracted, the distributed file dependency path parameters are accumulated to the node path, the resource performance dependency weight value is generated, and it is filled into the mapping relationship graph path set, the key path is optimally connected and screened, and a resource mapping optimization path set is generated; S3: Based on resource mapping, the path set is optimized, the transmission delay of the virtual machine and the bandwidth usage of the load balancer are analyzed, the memory idle rate is matched, the transmission efficiency distribution value is recursively adjusted, the transmission priority is reallocated to the nodes that exceed the critical point, and the iterative adjustment is performed to generate a resource priority migration sequence table; S4: Based on the resource priority migration sequence table, the target virtual machine operating status and network latency are updated in real time, memory usage and read / write rates are compared, resources are expanded for nodes whose difference values ​​exceed the range, the target node performance offset value is corrected to within the standard range, and a migration completion node performance status table is generated. The node operating load is monitored and adjusted to generate migration completion monitoring results.

[0019] The preliminary resource optimization data set includes processing capacity indicators, storage weighted indicators, and network delay indicators. The resource performance dependency weight values ​​include node performance differences and dependency path weights. The resource mapping optimization path set includes mapping relationship graph filling and path optimization results. The resource priority migration sequence table includes transmission efficiency adjustment and transmission priority setting. The migration completion node performance status table includes performance status update and performance standard correction. The migration completion monitoring results include monitoring data analysis and load adjustment feedback.

[0020] See also Figure 2 ,The specific steps for obtaining the preliminary resource optimization data set are: S111: Proportionally allocate the processing capacity of the virtual machine instance based on the processing capacity of the virtual machine instance, the block storage read and write rate, and the network delay time, and generate a processing capacity weight value of the virtual machine instance by comparing the load conditions and required resources of the differentiated virtual machine instances; First, the actual processing capacity of each virtual machine instance needs to be measured in detail, involving parameters such as the CPU utilization, memory occupancy, disk I / O rate, and network bandwidth usage of the monitoring virtual machine. Through the comprehensive evaluation of indicators, the total processing capacity value of each virtual machine instance can be obtained. The measurement method can use resource monitoring tools (such as the top command or the vmstat tool) to obtain the resource occupancy in real time and record it to obtain the total processing capacity of all virtual machine instances. By calculating the ratio of the processing capacity value of each virtual machine to the total processing capacity value, the processing capacity ratio of each virtual machine instance is obtained. For accuracy, the actual resource usage of the virtual machine instance must be considered during the calculation, and multiple sampling is required. The processing capacity ratio of each virtual machine instance will be used as the basis for the proportion of the instance in resource allocation. It is also necessary to set a certain threshold to identify instances with weak or overloaded processing capacity to ensure the balance and stability of resource allocation. It will be adjusted according to the priority or task characteristics of the virtual machine instance to ultimately ensure the optimal configuration of resources. Finally, the processing capacity weight value of the virtual machine instance is generated, which will be used to guide the dynamic scheduling and load balancing strategy of virtual machine resources.

[0021] S112: Based on the processing capacity weight value of the virtual machine instance, a weighted average calculation is performed on the block storage read / write rate and the network delay time, using the formula: ; Generate weighted block storage and network latency weight values; in, is the weighted average, is the weight coefficient of the block storage read and write rate, is the weight coefficient of network delay time, is the block storage read and write rate, is the network delay time; The benefit of the formula is that, by taking the weighted average of the block storage read and write rate and the network latency, the impact of these two parameters on the performance of the virtual machine can be considered at the same time, ensuring the reasonable allocation of resources. First, you need to obtain the block storage read and write rate and network latency The actual value of the block storage read and write rate can be obtained through hardware monitoring tools or the API provided by the cloud platform. The network delay time is measured by network testing tools (such as the ping command) or obtained through the network monitoring interface of the cloud platform. After monitoring and calculation, the block storage read and write rate is obtained. , network delay time , set the weight coefficient and ,These two coefficients are obtained through the performance test and historical data analysis of ,virtual machines; According to the formula, substitute the values ​​and calculate: ; First, calculate the block storage read and write rate part: ; Then calculate the network delay part: ; Next, add these two parts together and divide by the sum of the weight coefficients: ; The result shows that the comprehensive weight value obtained by weighted calculation is 28002.1213, which represents the comprehensive impact of block storage and network delay. This weight value will be used for priority sorting and resource scheduling in subsequent virtual machine resource allocation to ensure the optimal configuration and utilization of resources.

[0022] S113: Integrate the weighted block storage and network delay weight values ​​and the processing capacity weight into a two-dimensional matrix, sort according to the weighted results, screen qualified virtual machine instances, remove instances below the threshold, and generate a preliminary resource optimization data set; First, for each virtual machine instance, a two-dimensional matrix is ​​generated based on its processing power weight and the weighted storage and network delay weights. Each row of the matrix represents the relevant data of a virtual machine instance, where the first column represents the processing power weight of the virtual machine instance, and the second column represents the corresponding block storage and network delay weighted values. The processing power weight of the virtual machine instance is obtained by the above calculation, and the block storage and network delay weighted values ​​are obtained by weighting the weights in the formula. After the matrix is ​​generated, each virtual machine instance needs to be sorted. The sorting is performed from large to small according to the comprehensive weight value to ensure that resource allocation gives priority to meeting the needs of high-performance virtual machine instances. After the sorting is completed, virtual machine instances with strong processing power and excellent performance are screened according to the set threshold, and more resources are allocated to them. During the screening, the threshold can be adjusted according to the priority, task nature and resource occupancy of the virtual machine instance to achieve a more flexible resource scheduling effect. A preliminary resource optimization data set is generated based on the screening results. The data set will contain the processing power weight and weighted storage and network delay weights of each virtual machine instance, as well as resource allocation suggestions, as the basis for subsequent virtual machine resource scheduling decisions.

[0023] See also Figure 3 ,The specific steps for obtaining the resource performance dependency weight value are: S211: Based on the preliminary resource optimization data set, extract the performance indicator data between the virtual machine and the private subnet, perform a ratio analysis on the network traffic of the virtual machine and the subnet traffic, analyze the difference between the traffic ratio and the throughput and delay parameters, and generate a performance weight difference value between the virtual machine and the private subnet; The input and output traffic of the virtual machine network traffic are collected time period by time, and the network performance monitoring tool is called to record the throughput data. The throughput volatility is further analyzed to obtain the stable throughput value of each time period. For the acquisition of delay parameters, the delay value is calculated according to the time segment through multi-point monitoring of the network delay, and its segment mean is extracted. The throughput and delay parameters are processed in a normalized manner and unified to the same scale. Then, according to the relative changes between the throughput and delay parameters, the difference value matrix is ​​calculated, and the main difference parameters are selected in the difference value matrix for weight allocation. The weighted results of each difference value are accumulated item by item to form the performance weight difference value between the virtual machine and the private subnet. The performance difference change trend of the virtual machine in different time segments is further compared.

[0024] S212: Based on the performance weight difference between the virtual machine and the private subnet, call the path parameter set of the distributed file, identify the node path weight and bandwidth ratio in the path, superimpose the impact of the number of path dependencies, and use the formula: ; Calculate the node path dependency weight; in, represents the node path dependency weight, Indicates the path The path weight of each node, Indicates the path The bandwidth share of each node, represents the cumulative number of path dependencies, Represents the total number of nodes in the path, is the weight adjustment coefficient; The formula is useful because it introduces the path dependence times and weight adjustment factor , which can quantify the dependencies between path nodes and combine them with bandwidth share Reflects the contribution of the node to the overall path performance, thereby improving the flexibility and applicability of path weight calculation; Set the number of path nodes , the path weight of each node , , , , , node bandwidth ratio , , , , , the number of path dependencies , , , , , weight adjustment coefficient , the total number of nodes on the path ; Step 1: Calculate the comprehensive weight of each node and substitute it into the formula: Node 1: ; Node 2: ; Node 3, node 4, and node 5 are calculated in sequence , , ; Step 2: Accumulate the comprehensive weight of all nodes: ; Step 3: Calculate node path dependency weights: ; The result shows that the node path dependency weight value is 1.6298, which reflects the comprehensive influence of the nodes in the path in terms of bandwidth share, path dependency times, etc. This value can be used to further analyze the path performance.

[0025] S213: Based on the node path dependency weight, combined with the performance weight difference value between the virtual machine and the private subnet, the weight data of the two are normalized, the weight dependency of the overall resource performance is analyzed, and the resource performance dependency weight value is generated; The node path weight values ​​and the performance weight difference values ​​of the virtual machine and the private subnet are normalized, and the two weight values ​​are represented in a matrix according to the weight and difference dimensions. By comparing the corresponding weight difference distribution in the matrix, the normalized values ​​of the weight distribution of the virtual machine and the private subnet in each path node are calculated, and the normalized values ​​are accumulated to generate the resource performance dependency weight values. The degree of dependence on resource performance is classified according to the above weight values, and the resource performance dependency weight values ​​are organized into the final structured data.

[0026] See also Figure 4 , the specific steps for obtaining the resource mapping optimization path set are: S221: extracting each performance parameter in the resource set based on the resource performance dependency weight value, determining the weight difference of each pair of resources by comparing the performance values ​​between the resources, and classifying and counting the weight differences to generate a resource weight mapping node set; When extracting each performance parameter in the resource collection, it is grouped and processed according to the category of the resource performance parameter. The value of each group of performance parameters is extracted by calling the resource performance monitoring interface one by one, and a performance value set is generated in sequence. In the process of comparing the performance values ​​between resources, the weight value of the dependent performance parameter is used as the comparison basis, and the weight difference of each pair of performance parameters between resources is calculated in sequence. The calculation method is completed by setting the relative weight value formula of each group of performance parameters, and the performance difference results are recorded. In the classification statistics, resources with similar performance difference values ​​are classified into the same category to generate a weight difference statistical matrix. Each row represents the performance parameter range of a category of resources, and each column records the weight difference results of resources belonging to this category. Through the results recorded in the difference statistical matrix, the weight differences between resources are classified to generate a weight mapping node set. Each node in the node set is used to represent the mapping relationship between the dependency weights of resources, and the weight connection points between resources are numbered, which can be used in subsequent path screening and optimization processes to support weight association analysis between resources.

[0027] S222: Mapping a node set based on resource weights, by screening node and path weight combinations, analyzing the accumulated path weight values ​​one by one, calculating the dependency strength of the associated resource node combination, matching the calculation results with the weight values, and generating a key resource path set; By extracting the node weight values ​​of the resource weight mapping node set, the connection path of each node is analyzed in turn, and the dependency strength analysis model is set according to the cumulative value of the connection weight between nodes. When screening the node and path combination, nodes with larger weight values ​​are given priority, and in the combined path, path segments with similar weights are gradually screened. The calculation basis of the path cumulative value is the item-by-item superposition result of the connection node weights. When calculating the dependency strength, the cumulative total value and average value of the node weights of the path are analyzed one by one. For the path segments with significant weight differences between nodes, the weight ratios of the associated resource nodes are further calculated to determine the dependency strength of the node combination. After mapping the measurement results to the weight value interval, a key resource path set is generated according to the weight value. The resource path with higher weight is taken as the key path given priority. The paths in the set are sorted from high to low according to the weight value, providing basic support for subsequent path optimization.

[0028] S223: Based on the key resource path set, by comparing the dependency weight relationship of nodes in the path, the path segment with the optimal weight value is selected and connected, the integrity of the connected path is verified and the structure is adjusted, the optimized path is added to the mapping set, and the resource mapping optimized path set is obtained; When comparing the dependency weight relationship of nodes in the path, each path in the key resource path set is extracted one by one. By splitting the path, the weight value of each node in the path is recorded, and the weight difference of adjacent nodes in the path segment is calculated. When screening the optimal path segment, the path segment with smaller weight difference and higher cumulative value is given priority. A more stable path combination is formed through connection processing. When verifying the integrity of the connected path, the total weight value of the path and the uniformity of the node weight distribution are calculated to ensure that the path structure can meet the integrity requirements of resource dependency relationships in cross-cloud resource migration. In the process of adjusting the path structure, the path segments with large weight differences are split and recombined until the path weight distribution tends to be uniform. After the optimized path is added to the mapping set, the resource mapping optimized path set is obtained, which will be used as the core dependency structure of cross-cloud resource migration to guide the scheduling and path planning of resource migration and achieve efficient deployment and migration of resources.

[0029] See also Figure 5 , the specific steps for obtaining the resource priority migration sequence table are: S311: Optimize the path set based on resource mapping, extract the virtual machine transmission delay data, calculate the average delay value of each path, and perform coupling analysis with the bandwidth occupancy value of the subnet load balancer to evaluate the transmission delay characteristics, select the path that exceeds the target threshold, record the bandwidth occupancy ratio, and generate the analysis results of transmission delay and bandwidth occupancy; Collect the transmission delay data from the virtual machine to the subnet, record the delay time of each path through multiple measurements, obtain a continuous data set of delay time, divide the data into multiple time periods and calculate the average delay value of each path in each time period, and further calculate the fluctuation range of the delay data of each path, identify the path with large fluctuations as the target path that needs further analysis, measure the bandwidth occupancy of the target path, record the bandwidth value allocated to the path by the subnet load balancer and its actual usage ratio, and evaluate the transmission efficiency of the virtual machine on different paths based on the difference in the ratio of actual bandwidth occupancy to the allocated bandwidth value in combination with the delay time data, and calculate the transmission delay characteristic value of each path. Generate the analysis results of transmission delay and bandwidth occupancy based on the delay characteristic value and bandwidth occupancy ratio, which provides input data and evaluation basis for subsequent path optimization and priority allocation.

[0030] S312: Based on the analysis results of transmission delay and bandwidth occupancy, the relationship between path transmission load and memory occupancy is evaluated, and the transmission efficiency distribution value is recursively adjusted using the formula: ; generating a distribution adjustment result of path transmission efficiency; in, represents the transmission efficiency distribution value, For the The memory idle rate of the path, is the path bandwidth occupancy weight, is the path load value, is the adjustment factor, is the sum of the path transmission priority weights, Indicates the total number of paths; The formula is beneficial in that it refines the analysis of resource utilization between paths by combining the memory idle rate and the path bandwidth occupancy weight, and introduces the absolute value of the difference in path load values. Combined with the square root operation, it weakens the impact of path load anomalies on the calculation results, making the evaluation results of transmission efficiency distribution smoother. Set the total number of paths , the memory idle rate of the first path , the memory idle rate of the second path , the memory idle rate of the third path , the memory idle rate of the fourth path , the corresponding bandwidth occupation weight is , , , , the path load value is , , , , adjustment coefficient , the sum of path priorities ; Calculate the composite value for each path: Path 1: ; Path 2: ; Path 3: ; Path 4: ; Add up the comprehensive values ​​and calculate the transmission efficiency distribution value: ; This result shows that the transmission efficiency distribution value is , reflects the comprehensive transmission performance of the current path under the conditions of bandwidth utilization, memory idle rate and load distribution. This value can be used for the next transmission priority adjustment.

[0031] S313: based on the distribution adjustment result of the path transmission efficiency, re-assign priorities to the path nodes whose transmission efficiency exceeds the critical point, recursively adjust the node weights item by item according to the priorities, sort the nodes according to their criticality and load priority, determine the migration priority, and generate a resource priority migration sequence table; Select the path nodes whose transmission efficiency is lower than the distribution threshold, calculate the difference between the load of the inefficient path and the load of the remaining nodes item by item, reallocate the transmission priority of the inefficient path according to the load difference value, and adjust the path weight of the node according to the reallocated priority. By gradually iterating and optimizing the load distribution and priority distribution, all paths are finally prioritized and the migration priority sequence is output. The paths with higher transmission efficiency distribution are prioritized at the front of the sequence list, and a resource priority migration sequence list is generated based on the sorting results.

[0032] See also Figure 6 , the specific steps for obtaining the performance status table of the migration-completed node are: S411: Based on the resource priority migration sequence table, monitor the running status of the target virtual machine in real time, collect the network delay value and CPU occupancy rate of the target virtual machine, record the change data in combination with the node resource index, normalize the network delay and CPU occupancy rate, and obtain the real-time performance status of the target virtual machine; The monitoring data of network delay and CPU occupancy are collected and recorded, and the monitoring process is divided into multiple time segments to ensure data integrity and continuity of time series, so as to accurately capture the trend of operation status changing over time. For the collected network delay data, the average delay value in each time segment is calculated, and the delay fluctuation amplitude is calculated at the same time to identify the time segments with large fluctuations. In CPU occupancy monitoring, the occupancy changes in different time segments are also calculated, and the maximum and minimum occupancy values ​​of each time segment are recorded. The CPU load stability is further analyzed, and the mean data of network delay and CPU occupancy are normalized to make the two have a unified scale for subsequent calculation and analysis. The normalized network delay and CPU occupancy data are compared with the predefined target performance standard values, the performance offset value of each time segment is calculated, and the distribution trend of the time segment offset value is recorded. The performance status change law of the target virtual machine in different time segments is sorted out, and the real-time performance status of the target virtual machine is generated to provide basic data support for subsequent performance correction and resource expansion adjustment.

[0033] S412: Based on the real-time performance status of the target virtual machine, compare the memory usage rate and the read / write rate, determine the change in the mean memory usage through the difference value, calculate the performance deviation based on the fluctuation range of the read / write rate, correct the difference value, and adjust the resource expansion parameters using the formula: ; Generate a performance deviation correction result of the target virtual machine; in, Indicates the performance deviation value, is the current memory usage, is the target memory usage, is the current read and write rate, is the target read and write rate; The formula is useful in that it evaluates the deviation of memory usage and read / write rate at the same time. By combining square operation and smoothing, it avoids misjudgment caused by fluctuation of a single parameter, and further improves the robustness and accuracy of performance deviation calculation. Set the current memory usage of the target virtual machine , target memory usage , current read and write rate , target read and write rate ; The first step is to calculate the square of the difference in memory usage: ; The second step is to calculate the absolute value of the difference in read and write rates: ; The third step is to substitute the calculated results into the formula: ; This result shows that the performance deviation value , indicating that there is a certain deviation in the memory usage and read / write rate of the current virtual machine, and further analysis and adjustment of resource expansion parameters are needed to reduce the performance deviation value.

[0034] S413: Based on the performance deviation correction result of the target virtual machine, perform resource expansion to adjust the nodes that exceed the standard range, correct the real-time performance status of the nodes after expansion, normalize and adjust the performance values ​​to the standard range, summarize the node performance correction status, and generate a migration completion node performance status table; Dynamic resource adjustment is performed on all nodes whose difference values ​​exceed the standard range. First, the memory usage and read / write rates of the nodes that exceed the standard range are detected, and the time segments with the largest difference values ​​are recorded respectively. Resource expansion operations are performed for the performance data corresponding to the time segments. The memory usage deviation value is reduced by allocating more memory resources. At the same time, the read / write rate is adjusted by optimizing disk performance or increasing network bandwidth so that the read / write rate gradually approaches the target value. For the resource expansion operation of each node, the performance deviation value change trend is monitored in real time. The current performance deviation value is recalculated after each round of adjustment operation, and the performance indicators of the nodes are gradually optimized. When the performance deviation values ​​of all nodes return to the standard range, the performance status of the nodes is normalized, the performance status data of the corrected nodes are sorted, and the performance status records of all nodes are summarized into a performance status table of the migrated nodes, providing data support for subsequent resource configuration and migration strategy optimization.

[0035] See also Figure 7 , the specific steps to obtain the monitoring results after the migration is completed are: S421: Based on the migration completed node performance status table, extract node performance parameter data, read and filter the operation indicators node by node, perform time series analysis on CPU occupancy, memory usage, and network traffic values, and generate a node performance status record table; When extracting node performance parameter data, the performance parameters of each node that has been migrated are read in turn. The performance monitoring interface is called one by one according to the record order in the node performance status table to collect the core operation indicators such as CPU occupancy, memory usage, and network traffic of each node. The collected performance parameters are sorted by timestamp index, and the performance data of the node is classified and sorted according to the time series. After reading the operation indicators node by node, the change trend of each performance parameter is calculated based on the time series analysis method to determine the indicator fluctuation range and abnormal peak value. In the analysis, in the monitoring record of CPU occupancy, every 10 minutes is used as a sampling interval to calculate the average occupancy value and filter out abnormal points that exceed the preset threshold. For memory usage, the instantaneous peak value and continuous occupancy status are recorded according to the memory allocation strategy of the node running task. The network traffic data is counted separately according to the upload and download components to generate a time series curve of the network traffic of each node. By sorting and analyzing the data of each node, a node performance status record table is generated, which contains the performance indicator time series data and abnormal point marks of each node, providing a basic basis for subsequent load analysis and adjustment.

[0036] S422: Based on the node performance status record table, analyze the node load occupancy, mark and classify the nodes that exceed the load threshold, transfer the high-load node tasks to the low-load nodes, evaluate the low-load node performance indicators, verify the adjusted node load parameters, and obtain the node load adjustment results; According to the data in the node performance status record table, the load occupancy of each node is analyzed, and the CPU occupancy rate, memory usage rate and network traffic of the node are comprehensively scored to calculate the current load status of the node. The nodes whose load score values ​​exceed the threshold are marked as high-load nodes and classified and recorded. After the marking is completed, by selecting low-load nodes as the target nodes for task transfer, some running tasks on high-load nodes are dynamically allocated to low-load nodes. During the transfer process, the performance indicators of low-load nodes are evaluated one by one, especially focusing on the immediate change values ​​of CPU and memory after task transfer and the adjustment of network throughput to ensure that the low-load node is still within the controllable load range after receiving the task. When verifying the adjusted node load parameters, the CPU occupancy rate and memory usage rate of the node after task transfer are collected in real time by calling the monitoring interface again, and the load change trend is recorded to ensure that the adjusted node operation parameters meet the performance standards of the migration task. Finally, the node load adjustment results are obtained, and the performance parameters of the nodes before and after the load adjustment are archived to lay the foundation for the optimization of resource operation after migration.

[0037] S423: Based on the node load adjustment result, monitor the adjusted nodes in real time, capture the running load and performance parameter data, compare the load stability and performance item by item, record the dynamic information of the node after migration, summarize the monitoring data, and generate the migration completion monitoring result; When real-time monitoring is performed on the adjusted nodes, the load parameters and performance data of the node operation are continuously collected, and core indicators such as CPU occupancy, memory usage, and network traffic are recorded at the minute level. The dynamic changes of various parameters are captured in real time. When comparing the load stability item by item, the node parameters in the load adjustment results are used as the benchmark, and the current monitoring data is compared with the adjustment results to check whether the load changes tend to be stable. Focus on the long-term impact of the migration task on the node performance. In the comparison of performance, the running task completion time, response delay and resource occupancy of each node are recorded, and a performance fluctuation curve is generated to identify the time period of abnormal fluctuations and the corresponding load change data. When recording the dynamic information after the node migration, each performance parameter of the node is statistically divided into time segments, and classified and sorted to form a historical trend of performance. After the monitoring data is summarized, the load changes of each node before and after the migration task are analyzed to generate the migration completion monitoring results. The monitoring results include the node load adjustment, performance change trend and task migration effect evaluation.

[0038] The cross-cloud resource migration system is used to execute the above cross-cloud resource migration method, and the system includes: The node screening module calculates the weighted value of the block storage read / write rate and the network delay time based on the processing capacity of the virtual machine instance, the block storage read / write rate, and the network delay time, combines the three parameters in proportion for evaluation, and sorts the nodes, screening the nodes with scores greater than the set threshold to obtain a set of high-scoring nodes; The path dependency module calculates the weight of the node attributes of the distributed file dependency path based on the high-score node set, accumulates the file path length to the node weight value, selects the path based on the performance difference between the virtual machine and the subnet, and generates the path weight optimization result; The transmission priority module analyzes the transmission delay and bandwidth occupancy ratio of the node based on the path weight optimization results, re-analyzes the transmission efficiency distribution in combination with the node memory idle rate, adjusts the priority of the node whose occupancy rate exceeds the threshold, and generates an optimized transmission path sequence; The migration performance monitoring module monitors the target virtual machine's operating status and network latency in real time based on the optimized transmission path sequence, compares the node memory usage and block storage read and write rate differences, corrects performance offsets and records performance status, and generates migration completion monitoring results.

[0039] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A cross-cloud resource migration method, characterized in that: The following steps are involved: S1: Based on the processing capacity of the virtual machine instances, the block storage read / write rate, and the network delay time, the processing capacity of the virtual machine instances is proportionally allocated, the block storage read / write rate and the network delay time are weighted averaged, integrated into a two-dimensional matrix, sorted and removed low-threshold nodes, and a preliminary resource optimization data set is generated; S2: Based on the preliminary resource optimization data set, extract the performance weight difference between the virtual machine and the private subnet, accumulate the distributed file dependency path parameters to the node path, generate the resource performance dependency weight value, and fill it into the mapping relationship graph path set, perform optimal connection screening on the key path, and generate the resource mapping optimization path set; S3: Based on the resource mapping optimization path set, the transmission delay of the virtual machine and the bandwidth occupation of the load balancer are analyzed, the memory idle rate is matched, the transmission efficiency distribution value is recursively adjusted, the transmission priority is reallocated to the nodes exceeding the critical point, the transmission priority is iteratively adjusted, and a resource priority migration sequence table is generated; S4: Based on the resource priority migration sequence table, the target virtual machine operating status and network delay are updated in real time, the memory usage and read and write rates are compared, resources are expanded for nodes whose difference values ​​exceed the range, the target node performance offset value is corrected to within the standard range, and a migration completion node performance status table is generated. The node operating load is monitored and adjusted to generate a migration completion monitoring result.

2. The cross-cloud resource migration method according to claim 1, characterized in that: The steps for obtaining the preliminary resource optimization data set are specifically as follows: S111: Proportionally allocate the processing capacity of the virtual machine instance based on the processing capacity of the virtual machine instance, the block storage read and write rate, and the network delay time, and generate a processing capacity weight value of the virtual machine instance by comparing the load conditions and required resources of the differentiated virtual machine instances; S112: Based on the processing capacity weight value of the virtual machine instance, a weighted average calculation is performed on the block storage read / write rate and the network delay time, using the formula: ； Generate weighted block storage and network latency weight values; in, is the weighted average, is the weight coefficient of the block storage read and write rate, is the weight coefficient of network delay time, is the block storage read and write rate, is the network delay time; S113: Integrate the weighted block storage and network delay weight values ​​and the processing capacity weight into a two-dimensional matrix, sort according to the weighted results, screen qualified virtual machine instances, remove instances below the threshold, and generate a preliminary resource optimization data set.

3. The cross-cloud resource migration method according to claim 2, characterized in that: The steps for obtaining the resource performance dependency weight value are specifically as follows: S211: Based on the preliminary resource optimization data set, extract the performance indicator data between the virtual machine and the private subnet, perform a ratio analysis on the network traffic of the virtual machine and the subnet traffic, analyze the difference between the traffic ratio and the throughput and delay parameters, and generate a performance weight difference value between the virtual machine and the private subnet; S212: Based on the performance weight difference between the virtual machine and the private subnet, call the path parameter set of the distributed file, identify the node path weight and bandwidth ratio in the path, superimpose the influence of the number of path dependencies, and use the formula: ； Calculate the node path dependency weight; in, represents the node path dependency weight, Indicates the path The path weight of each node, Indicates the path The bandwidth share of each node, represents the cumulative number of path dependencies, Represents the total number of nodes in the path, is the weight adjustment coefficient; S213: Based on the node path dependency weight, combined with the performance weight difference value between the virtual machine and the private subnet, the weight data of the two are normalized, the weight dependency of the overall resource performance is analyzed, and the resource performance dependency weight value is generated.

4. The cross-cloud resource migration method according to claim 3, characterized in that: The steps for obtaining the resource mapping optimization path set are specifically as follows: S221: extracting each performance parameter in the resource set based on the resource performance dependency weight value, determining the weight difference of each pair of resources by comparing the performance values ​​between the resources, and classifying and counting the weight differences to generate a resource weight mapping node set; S222: Based on the resource weight mapping node set, by screening the node and path weight combination, analyzing the path weight accumulation value one by one, calculating the dependency strength of the associated resource node combination, matching the calculation result with the weight value, and generating a key resource path set; S223: Based on the key resource path set, by comparing the dependency weight relationship of the nodes in the path, the path segment with the optimal weight value is screened and connected, the integrity of the connected path is verified and the structure is adjusted, the optimized path is added to the mapping set, and the resource mapping optimized path set is obtained.

5. The cross-cloud resource migration method according to claim 4, characterized in that: The steps for obtaining the resource priority migration sequence table are specifically as follows: S311: based on the resource mapping optimization path set, extract the virtual machine transmission delay data, calculate the average delay value of each path, and perform coupling analysis with the subnet load balancer bandwidth occupancy value, evaluate the transmission delay characteristics, select the path exceeding the target threshold, record the bandwidth occupancy ratio, and generate the analysis results of transmission delay and bandwidth occupancy; S312: Based on the analysis results of the transmission delay and bandwidth occupancy, the relationship between the path transmission load and the memory occupancy is evaluated, and the transmission efficiency distribution value is recursively adjusted using the formula: ； generating a distribution adjustment result of path transmission efficiency; in, represents the transmission efficiency distribution value, For the The memory idle rate of the path, is the path bandwidth occupancy weight, is the path load value, is the adjustment factor, is the sum of the path transmission priority weights, Indicates the total number of paths; S313: Based on the distribution adjustment result of the path transmission efficiency, reallocate priorities to the path nodes whose transmission efficiency exceeds the critical point, recursively adjust the node weights item by item according to the priority, sort the nodes according to their criticality and load priority, determine the migration priority, and generate a resource priority migration sequence table.

6. The cross-cloud resource migration method according to claim 5, characterized in that: The steps for obtaining the migration completed node performance status table are specifically as follows: S411: Based on the resource priority migration sequence table, monitor the running status of the target virtual machine in real time, collect the network delay value and CPU occupancy rate of the target virtual machine, record the change data in combination with the node resource index, normalize the network delay and CPU occupancy rate, and obtain the real-time performance status of the target virtual machine; S412: Based on the real-time performance status of the target virtual machine, compare the memory usage rate and the read / write rate, determine the change in the mean memory usage through the difference value, calculate the performance deviation in combination with the read / write rate fluctuation range, correct the difference value, and adjust the resource expansion parameters using the formula: ； Generate a performance deviation correction result of the target virtual machine; in, Indicates the performance deviation value, is the current memory usage, is the target memory usage, is the current read and write rate, is the target read and write rate; S413: Based on the performance deviation correction result of the target virtual machine, perform resource expansion to adjust the nodes that exceed the standard range, correct the real-time performance status of the expanded nodes, normalize and adjust the performance values ​​to the standard range, summarize the node performance correction status, and generate a migration completed node performance status table.

7. The cross-cloud resource migration method according to claim 6, characterized in that: The steps for obtaining the migration completion monitoring result are specifically as follows: S421: Based on the migration completed node performance status table, extract node performance parameter data, read and filter the operation indicators node by node, perform time series analysis on CPU occupancy, memory usage, and network traffic values, and generate a node performance status record table; S422: Based on the node performance status record table, analyze the node load occupancy, mark and classify the nodes that exceed the load threshold, transfer the high-load node tasks to the low-load nodes, evaluate the low-load node performance indicators, verify the adjusted node load parameters, and obtain the node load adjustment results; S423: Based on the node load adjustment result, monitor the adjusted nodes in real time, capture the operating load and performance parameter data, compare the load stability and performance item by item, record the dynamic information of the node after migration, summarize the monitoring data, and generate the migration completion monitoring result.

8. A cross-cloud resource migration system, characterized in that: According to the cross-cloud resource migration method according to any one of claims 1 to 7, the system comprises: The node screening module calculates the weighted value of the block storage read / write rate and the network delay time based on the processing capacity of the virtual machine instance, the block storage read / write rate, and the network delay time, combines the three parameters in proportion for evaluation, and sorts the nodes, screening the nodes with scores greater than the set threshold to obtain a set of high-scoring nodes; The path dependency module calculates the weight of the distributed file dependency path node attributes based on the high-score node set, accumulates the file path length to the node weight value, selects the path based on the performance difference between the virtual machine and the subnet, and generates a path weight optimization result; The transmission priority module analyzes the transmission delay and bandwidth occupancy ratio of the node based on the path weight optimization result, re-analyzes the transmission efficiency distribution in combination with the node memory idle rate, adjusts the priority of the node whose occupancy rate exceeds the threshold, and generates an optimized transmission path sequence; The migration performance monitoring module monitors the target virtual machine's operating status and network delay in real time based on the optimized transmission path sequence, compares the node memory usage and block storage read and write rate difference, corrects performance offsets and records performance status, and generates migration completion monitoring results.

Images