A cloud data center virtual machine migration optimization method

Abstract

The application discloses a cloud data center virtual machine migration optimization method and relates to the technical field of data processing. The technical solution points of the method comprise the following steps: obtaining resource fragment data, virtual machine resource occupation data, idle resource distribution data and historical fragment arrangement records of a physical node of a cloud data center, and constructing a resource fragment data set; collecting fragment features in the resource fragment data set to obtain fragment feature data, and screening a high-reuse migration target node according to the fragment feature data to obtain a target node set; performing mode statistics on the historical fragment arrangement records to obtain fragment mode data, and constructing a resource reuse optimization database according to the fragment mode data; and the effect is that the virtual machine migration into scattered resource blocks avoids further segmentation of node resources, effectively controls the node resource fragmentation rate, and reserves continuous available resource space for subsequent virtual machine deployment.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically, to a method for optimizing virtual machine migration in cloud data centers. Background Technology

[0002] As the scale of business carried by cloud data centers continues to expand, the dynamic deployment and migration of virtual machines are becoming increasingly frequent, leading to the continuous fragmentation of physical node resources such as CPU, memory, and disk. Resource fragmentation manifests as a large number of scattered, idle resource blocks within a node that cannot be efficiently utilized by virtual machines. These fragments not only reduce the resource utilization rate of physical nodes but also increase the risk of virtual machine deployment failures, severely restricting the overall operational efficiency of cloud data centers.

[0003] Existing virtual machine migration strategies primarily focus on load balancing and resource capacity matching, only considering whether the total amount of idle resources on the target node meets the virtual machine's needs, without systematically considering the resource fragmentation state of the node. During the migration process, virtual machines often directly occupy fragmented idle resource blocks, further exacerbating the node's fragmentation. This makes it difficult for subsequently deployed virtual machines to find continuous idle resources, necessitating frequent additional defragmentation operations, increasing data center operational costs and migration latency.

[0004] Meanwhile, current migration optimization solutions lack effective utilization of historical fragmentation data, failing to extract fragmentation patterns from past operation and maintenance cases, and making it difficult to achieve intelligent generation of migration solutions. Although some solutions introduce fragmentation filtering mechanisms, they lack multi-dimensional adaptation verification and integration optimization processes, making it impossible to effectively assess the resource reuse efficiency and subsequent integration difficulty of the migration solution. This results in poor implementation of the generated migration solutions, with high fragmentation rates and operation and maintenance costs. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the present invention aims to provide a method for optimizing virtual machine migration in cloud data centers.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for optimizing virtual machine migration in a cloud data center, comprising the following steps:

[0008] Obtain resource fragment data, virtual machine resource usage data, idle resource distribution data, and historical fragmentation records of physical nodes in the cloud data center to construct a resource fragment dataset;

[0009] Fragment feature data is obtained by collecting fragment features from resource fragment datasets, and a set of target nodes is obtained by filtering high-reusability migration target nodes based on fragment feature data;

[0010] By performing pattern statistics on historical fragment records, fragment pattern data is obtained, and a resource reuse optimization database is constructed based on the fragment pattern data.

[0011] Input the resource requirements of the virtual machines to be migrated, the node fragmentation threshold, and the reuse efficiency indicators into the resource reuse optimization database to generate an initial migration and reuse plan.

[0012] Fragment adaptation verification is performed on the initial migration and reuse scheme to obtain the adaptation verification result;

[0013] Based on the adaptation verification results, the initial migration reuse scheme is integrated and optimized to obtain the final migration reuse scheme. The reuse effect of the final migration reuse scheme is verified to obtain the migration optimization results.

[0014] Preferably, the resource fragment dataset includes CPU fragment data, memory fragment data, disk fragment data, free resource block size, fragment distribution location, and fragment defragmentation time data.

[0015] Preferably, the target node set is obtained by filtering high-reusability migration target nodes based on fragment feature data, specifically including the following steps:

[0016] A node reuse evaluation matrix is ​​constructed based on fragment feature data, and the reuse evaluation matrix is ​​classified into levels to obtain reuse level results;

[0017] The physical node fragment data in the reuse level results are integrated to obtain the node fragment dataset;

[0018] The resource reuse degree is calculated based on the node fragment dataset to obtain the reuse degree result;

[0019] The target node set is formed based on the reuse rate results.

[0020] Preferably, the initial migration and reuse scheme is subjected to fragment adaptation verification to obtain the adaptation verification result, which specifically includes the following steps:

[0021] The initial migration and reuse scheme is marked with fragment region numbers to obtain region marking results, and the initial migration and reuse scheme is judged for fragment adaptation to obtain adaptation difference;

[0022] The reuse rate is analyzed based on the regional labeling results to obtain the reuse difference;

[0023] The difficulty value is obtained by assessing the difficulty of fragment integration of the initial migration and reuse scheme;

[0024] The adaptation verification result is determined based on the adaptation difference, reuse difference, and difficulty value.

[0025] Preferably, the initial migration and reuse scheme is subjected to fragment adaptation judgment to obtain the adaptation difference, which specifically includes the following steps:

[0026] Extract standard fragment adaptation benchmark data to obtain adaptation benchmark dataset;

[0027] Obtain the initial migration reuse scheme prediction adaptation data to obtain the prediction adaptation dataset;

[0028] The adaptation results are obtained by comparing the predicted adaptation dataset with the adaptation benchmark dataset.

[0029] The adaptation difference is obtained by calculating the difference between the adaptation results.

[0030] Preferably, the resource reuse rate is analyzed based on the area marking results to obtain the reuse difference, specifically including the following steps:

[0031] Obtain standard reuse rate data to obtain a standard reuse dataset;

[0032] Obtain the predicted reuse rate data of the initial migration reuse scheme to obtain the predicted reuse dataset;

[0033] The reuse results are obtained by comparing the predicted reuse dataset with the standard reuse dataset;

[0034] The reuse results are quantified to obtain the reuse difference.

[0035] Preferably, the difficulty value is obtained by assessing the fragment integration difficulty of the initial migration and reuse scheme, specifically including the following steps:

[0036] Obtain the fragment quantity data to get the fragment quantity value;

[0037] Obtain the dispersion value by acquiring fragment dispersion data;

[0038] The difficulty value is obtained by weighting the number of fragments and the dispersion value.

[0039] Preferably, the initial migration and reuse scheme is optimized by fragment integration based on the adaptation verification results to obtain the final migration and reuse scheme, specifically including the following steps:

[0040] The direction and priority of fragment integration are determined based on the adaptation verification results.

[0041] Based on the integration direction and priority, resource allocation, fragment merging strategies, and migration order are adjusted to obtain the final migration and reuse plan.

[0042] Preferably, the final migration reuse scheme is validated to obtain migration optimization results, specifically including the following steps:

[0043] The verification results are obtained by comparing the resource fragmentation rate after the final migration and reuse scheme is executed with the preset fragmentation threshold.

[0044] The verification results are compared with the preset fragmentation threshold.

[0045] If the verification result is less than or equal to the preset fragmentation threshold, then output the migration optimization qualified result;

[0046] If the verification result is greater than the preset judgment threshold, the fragment reorganization result will be output.

[0047] Compared with existing technologies, this invention has the following advantages: By constructing a resource fragment dataset containing CPU, memory, disk fragment data, and idle resource distribution information, the fragmentation state of physical nodes is comprehensively captured; simultaneously, in the target node selection stage, a reuse evaluation matrix is ​​constructed based on fragment feature data, prioritizing nodes with good continuity and concentrated distribution of idle resources as migration targets, avoiding further fragmentation of node resources by virtual machines migrating into scattered resource blocks, effectively controlling the node resource fragmentation rate, and reserving continuous and usable resource space for subsequent virtual machine deployment; by accumulating historical fragmentation patterns in the resource reuse optimization database, continuous idle resource blocks that are highly adapted to the resource requirements of virtual machines are matched; at the same time, through fragment adaptation verification, reuse difference analysis, and integration difficulty assessment, it is ensured that virtual machines can maximize the utilization of node idle resources after migration, reduce invalid resource occupation, thereby significantly improving resource block utilization, enabling node idle resources to be fully utilized, and effectively solving the problem of low resource utilization in the solution. Attached Figure Description

[0048] Figure 1 This is a schematic diagram illustrating the steps of a cloud data center virtual machine migration optimization method according to an embodiment of the present invention;

[0049] Figure 2 This is a schematic diagram illustrating the steps of obtaining the target node set in a cloud data center virtual machine migration optimization method provided in an embodiment of the present invention. Detailed Implementation

[0050] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0051] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0052] Secondly, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places throughout this specification does not necessarily refer to the same embodiment, nor is it a single embodiment or an embodiment selectively excluded from other embodiments.

[0053] Reference Figures 1-2 As shown.

[0054] Example 1 further illustrates the cloud data center virtual machine migration optimization method proposed in this invention.

[0055] A method for optimizing virtual machine migration in a cloud data center, comprising the following steps:

[0056] Obtain resource fragment data, virtual machine resource usage data, idle resource distribution data, and historical fragmentation records of physical nodes in the cloud data center to construct a resource fragment dataset;

[0057] Fragment feature data is obtained by collecting fragment features from resource fragment datasets, and a set of target nodes is obtained by filtering high-reusability migration target nodes based on fragment feature data;

[0058] By performing pattern statistics on historical fragment records, fragment pattern data is obtained, and a resource reuse optimization database is constructed based on the fragment pattern data.

[0059] Input the resource requirements of the virtual machines to be migrated, the node fragmentation threshold, and the reuse efficiency indicators into the resource reuse optimization database to generate an initial migration and reuse plan.

[0060] First, pattern statistics are performed on historical defragmentation records to obtain defragmentation pattern data. These records store detailed information about each defragmentation operation, including the CPU fragmentation rate, memory fragmentation rate, disk fragmentation rate, free resource block size distribution, and fragment location of the node before defragmentation, as well as the resource utilization, fragmentation rate, improvement, and defragmentation time of the node after defragmentation. The pattern statistics process extracts regular defragmentation experiences from massive amounts of data. For example, statistics show that when the average size of a node's free resource blocks matches the resource requirements of the virtual machine to be migrated to a certain proportion, the resource utilization of the node after defragmentation can increase by an average of over 30%; or when the fragmentation distribution concentration of a node is within a specific range, the defragmentation time is significantly reduced, and the probability of generating new fragments after defragmentation is much lower than in other ranges. These patterns mined from historical data constitute the fragmentation pattern data, describing the effectiveness and efficiency of defragmentation under different combinations of fragmentation characteristics and the possibility of subsequent resource reuse, providing empirical support for subsequent database construction.

[0061] A resource reuse optimization database is built based on fragmentation pattern data. The core of this database is to transform fragmentation pattern data into directly callable matching rules and decision models. The database stores the mapping relationship between various fragmentation feature combinations and the reuse effect of optimal defragmentation strategies, and also includes target node matching rules corresponding to different virtual machine resource requirements. For example, the database records that when the CPU requirement of the virtual machine to be migrated is 4 cores and the memory requirement is 8GB, the average size of the idle CPU resource blocks of the matching nodes is not less than 4 cores, the average size of the idle memory resource blocks is not less than 8GB, and the fragmentation concentration is higher than a preset threshold, the reuse efficiency can be optimized. The database also integrates defragmentation effect data under different node fragmentation thresholds. For example, when the CPU fragmentation rate of a node exceeds 20%, nodes with consecutive idle CPU resource blocks should be matched first to avoid further exacerbating fragmentation. The completed database is equivalent to accumulating all past experience in fragmentation defragmentation and virtual machine migration in cloud data centers, enabling rapid matching of optimal strategies when generating subsequent solutions without the need for complex repetitive calculations each time.

[0062] Subsequently, the resource requirements of the virtual machine to be migrated, including node fragmentation thresholds, reuse efficiency, and indicator inputs, are optimized in the resource reuse database to generate an initial migration and reuse plan. The resource requirements of the virtual machine to be migrated include the number of CPU cores, memory size, disk capacity, and corresponding performance requirements. The node fragmentation threshold is a preset upper limit for node fragmentation rate. When the fragmentation rate of the target node exceeds this threshold, fragmentation needs to be performed synchronously after migration. The reuse efficiency indicator is the target value for measuring the resource utilization efficiency of the migration plan. After receiving these input parameters, the database calls its internal fragmentation pattern data and matching rules to match nodes in the target node set for the virtual machine to be migrated and generates a preliminary migration and reuse plan. For example, if the resource requirements of the virtual machine to be migrated are 4 CPU cores and 8GB of memory, the node fragmentation threshold is 15%, and the reuse efficiency indicator is no less than 80%, the database will prioritize selecting nodes in the target node set whose idle CPU and memory resource block sizes perfectly match the virtual machine's requirements and whose current fragmentation rate is less than 15%. Simultaneously, referring to historical fragmentation pattern data, nodes with a post-migration reuse efficiency of over 80% are selected as target nodes to form the initial migration and reuse plan. This solution clearly defines the migration path of the target node of the virtual machine to be migrated and whether fragmentation needs to be performed simultaneously, providing a basic framework for subsequent verification and optimization.

[0063] To more accurately measure the reuse efficiency of the solution, the reuse efficiency calculation formula is: Reuse Efficiency = Utilization Rate of Idle Resource Blocks on the Node After Migration × Scraping Efficiency Correction Coefficient. Here, the utilization rate of idle resource blocks on the node after migration refers to the ratio of the size of the idle resource blocks directly occupied by the virtual machine after migration to the total size of the node's idle resources. The scrapping efficiency correction coefficient is calculated based on the node's historical scrapping time data; the shorter the historical time, the closer the coefficient value is to 1, and vice versa. For example, if the target node has 8 idle CPU cores, including two consecutive 4-core idle blocks, 16GB of idle memory, and one consecutive 8GB idle block, and the virtual machine to be migrated requires 4 CPU cores and 8GB of memory, and after migration, directly occupies the consecutive idle resource blocks, the idle resource block utilization rate is 75%. Simultaneously, if the node's historical scrapping time is short, with a scrapping efficiency correction coefficient of 0.95, then the reuse efficiency of this solution is approximately 71%, which meets the preset reuse efficiency target. Through such quantitative calculations, the resource reuse optimization database can quickly evaluate the reuse efficiency of different node matching schemes, select the initial migration reuse schemes that meet the indicator requirements, and ensure that the schemes meet the virtual machine resource requirements while maximizing resource reuse efficiency and reducing the additional overhead of defragmentation.

[0064] Fragment adaptation verification is performed on the initial migration and reuse scheme to obtain the adaptation verification result;

[0065] Based on the adaptation verification results, the initial migration reuse scheme is integrated and optimized to obtain the final migration reuse scheme. The reuse effect of the final migration reuse scheme is verified to obtain the migration optimization results.

[0066] The resource fragment dataset includes CPU fragment data, memory fragment data, disk fragment data, free resource block size, fragment distribution location, and fragment defragmentation time data.

[0067] Resource fragmentation datasets form the data foundation for virtual machine migration optimization methods in cloud data centers. They include CPU fragmentation data, memory fragmentation data, disk fragmentation data, free resource block size, fragment distribution location, and defragmentation time data. During the daily operation of a cloud data center, the CPU, memory, and disk resources of physical nodes are gradually fragmented into multiple discontinuous small blocks as virtual machines are created, deleted, and migrated, forming resource fragments. These fragments cannot be efficiently utilized by a single virtual machine, leading to decreased resource utilization. The construction of resource fragmentation datasets aims to comprehensively capture this fragmentation information. For example, CPU fragmentation data records the distribution of unused CPU cores or time slices on each physical node; memory fragmentation data reflects the size and location of scattered free pages in physical memory; and disk fragmentation data shows the discontinuous free sectors on the storage medium. Free resource block size and fragment distribution location data further clarify the scale of these fragments and their specific distribution on physical nodes, while defragmentation time data records the time spent on defragmentation operations in the past, providing a reference for subsequent optimization decisions. By integrating this multi-dimensional data, the resource fragmentation dataset can fully present the resource fragmentation status of each physical node, providing reliable data support for subsequent fragmentation feature collection, target node screening, and fragmentation pattern statistics. This allows the entire virtual machine migration optimization process to be based on comprehensive and accurate information, thereby effectively improving resource reuse efficiency and virtual machine migration results.

[0068] The target node set is obtained by filtering high-reusability migration target nodes based on fragment feature data, specifically including the following steps:

[0069] A node reuse evaluation matrix is ​​constructed based on fragment feature data, and the reuse evaluation matrix is ​​classified into levels to obtain reuse level results;

[0070] The physical node fragment data in the reuse level results are integrated to obtain the node fragment dataset;

[0071] The resource reuse degree is calculated based on the node fragment dataset to obtain the reuse degree result;

[0072] The target node set is formed based on the reuse rate results.

[0073] First, a node reuse evaluation matrix needs to be constructed based on fragmentation characteristic data. This matrix transforms the fragmentation characteristics of each physical node into quantitative evaluation dimensions. Each row corresponds to a physical node, and each column corresponds to a key fragmentation characteristic indicator. These indicators include CPU fragmentation rate, memory fragmentation rate, disk fragmentation rate, average size of free resource blocks, fragmentation distribution concentration, and historical defragmentation time. For example, CPU fragmentation rate is the proportion of unused CPU resources scattered into discontinuous small blocks out of the total free CPU resources; memory fragmentation rate reflects the proportion of discontinuous pages in the node's free memory; disk fragmentation rate corresponds to the proportion of discontinuous free sectors on the storage medium; average size of free resource blocks reflects the average size of a single free resource block on the node; fragmentation distribution concentration indicates the degree of clustering of free resource blocks on the node; and historical defragmentation time reflects the time cost of past defragmentation operations on the node. The completed node reuse evaluation matrix integrates the quantitative characteristics of all nodes, providing data for subsequent reuse level classification.

[0074] The reuse evaluation matrix is ​​graded to obtain reuse level results. Grading is based on the values ​​of various fragmentation characteristic indicators, classifying the reuse potential of each node. Reuse levels are typically divided into multiple tiers, with higher levels indicating that the node's fragmentation characteristics are more conducive to resource reuse. Specifically, nodes with lower CPU and memory fragmentation rates indicate more concentrated idle resources, making them easier for virtual machines to utilize, resulting in a higher reuse level. Nodes with larger average idle resource block sizes are better able to match the resource needs of virtual machines, leading to a higher reuse level. Higher fragmentation concentration means more concentrated idle resource blocks on the node, eliminating the need for virtual machines to occupy multiple fragments after migration, further improving reuse level. Nodes with shorter historical defragmentation times have lower defragmentation costs, making subsequent fragment consolidation more efficient, also resulting in a higher reuse level. Through this multi-indicator weighted or threshold-based method, each physical node is assigned a corresponding reuse level, forming reuse level results and initially screening nodes with high reuse potential.

[0075] Subsequently, the physical node fragment data from the reuse level results are integrated to obtain a node fragment dataset. The purpose of this step is to select nodes from all physical nodes that meet the preset reuse level criteria and extract all fragment feature data corresponding to these nodes to form a dedicated node fragment dataset. For example, if nodes with a preset reuse level of level two or higher are considered high-reuse-potential nodes, then all nodes with a reuse level lower than level two will be excluded, and only CPU fragment data, memory fragment data, disk fragment data, free resource block size, fragment distribution location, and fragmentation time data for nodes at level two or higher will be retained. The node fragment dataset no longer includes data from low-reuse-potential nodes, allowing for a more focused and accurate calculation of subsequent reuse levels and reducing the interference of invalid data on the evaluation results.

[0076] Calculating resource reuse based on node fragment datasets is the core quantitative step in the entire screening process. Resource reuse is a comprehensive indicator that measures the degree of matching between the idle resources of a node and the resource requirements of the virtual machine to be migrated. The calculation formula can be expressed as: Reuse = Matching Coefficient × Idle Resource Block Utilization × Fragmentation Efficiency Coefficient. The matching coefficient is the ratio of the total amount of idle resources on a node to the resource requirements of the virtual machine to be migrated. If the total amount of idle resources on the node is greater than or equal to the resource requirements of the virtual machine, the matching coefficient is 1; otherwise, it is proportionally adjusted. The idle resource block utilization rate is the ratio of the total size of idle resource blocks in the node that can be directly occupied by the virtual machine's resource requirements to the total amount of idle resources on the node. For example, if the virtual machine to be migrated requires 4 CPU cores and 8GB of memory, and the node has two consecutive 4-core CPU idle blocks and one consecutive 8GB memory idle block, the proportion of the total size of these idle blocks to the total amount of idle resources on the node is the idle resource block utilization rate. The defragmentation efficiency coefficient is a correction coefficient calculated based on the node's historical defragmentation time. The shorter the historical defragmentation time, the closer the coefficient value is to 1; the longer the time, the lower the coefficient value, thus reflecting the impact of node defragmentation cost on reuse. Through this formula, each node will obtain a specific reuse value, forming a reuse result. The higher the value, the more suitable the node is as a migration target.

[0077] Finally, a target node set is formed based on the reuse rate results. A reuse rate threshold is set during the selection process; only nodes with a reuse rate value that reaches or exceeds this threshold are included in the target node set. For example, if the reuse rate threshold is set to 0.7, then all nodes with a reuse rate greater than or equal to 0.7 will be retained, forming the final target node set. These nodes not only possess a high reuse level but also, through quantitative verification of resource reuse, can maximize the integration of idle resources while meeting virtual machine resource requirements, reducing the generation of new fragmentation after migration. This provides a reliable selection of target nodes for subsequently generating efficient virtual machine migration and reuse solutions.

[0078] The initial migration and reuse scheme is subjected to fragment adaptation verification to obtain the adaptation verification result, which specifically includes the following steps:

[0079] The initial migration and reuse scheme is marked with fragment region numbers to obtain region marking results, and the initial migration and reuse scheme is judged for fragment adaptation to obtain adaptation difference;

[0080] The reuse rate is analyzed based on the regional labeling results to obtain the reuse difference;

[0081] The difficulty value is obtained by assessing the difficulty of fragment integration of the initial migration and reuse scheme;

[0082] The adaptation verification result is determined based on the adaptation difference, reuse difference, and difficulty value.

[0083] First, the initial migration and reuse scheme is marked with fragmented regions to obtain region marking results. Simultaneously, fragment adaptation is performed on the initial migration and reuse scheme to obtain adaptation differences. The core of fragmented region marking is to number all free resource blocks of the target node according to their physical location and contiguousness. For example, consecutive CPU free resource blocks are marked as CPU1 and CPU2, consecutive memory free resource blocks are marked as MEM1 and MEM2, and disk free resource blocks are marked as DISK1 and DISK2, clearly presenting the location and size of each free resource block. Adaptation judgment involves matching the resource requirements of the virtual machine to be migrated with these marked fragmented regions and calculating the adaptation difference. The formula for calculating the adaptation difference is: Adaptation difference = Total virtual machine resource requirements - Total size of successfully matched consecutive free resource blocks. For example, if the virtual machine to be migrated requires 4 CPU cores and 8GB of memory, and the target node has a contiguous block of 2 CPU cores and a contiguous block of 2 CPU cores, as well as a contiguous block of 6GB of memory and a contiguous block of 2GB of memory, and the total size of the successfully matched contiguous free resource blocks is 4 CPU cores and 6GB of memory, then the adaptation difference for the CPU part is 0, and the adaptation difference for the memory part is 2GB. The overall adaptation difference reflects the part of the virtual machine's resource requirements that cannot be directly met by the contiguous free resource blocks. The larger the difference, the more fragmented resources the solution needs to occupy, and the worse the adaptability.

[0084] Next, the resource reuse rate is analyzed based on the region marking results to obtain the reuse difference. The resource reuse rate analysis is based on the region marking results, calculating the actual utilization of idle resource blocks on the target node after virtual machine migration, and comparing it with the preset reuse efficiency index to obtain the reuse difference. The formula for calculating the reuse difference is: Reuse Difference = Preset Reuse Efficiency Index - Estimated Reuse Efficiency. Where the estimated reuse efficiency = Total size of consecutive idle resource blocks occupied by virtual machines ÷ Total idle resource size of the node. For example, if the preset reuse efficiency index is 80%, the target node has 8 idle CPU cores, of which 4 cores are continuously occupied by virtual machines, and a total idle memory of 16GB, of which 6GB is continuously occupied by virtual machines, the estimated reuse efficiency = (4+6) ÷ (8+16) ≈ 0.42, which is approximately 42%. Therefore, the reuse difference is 80% minus 41% (38%). A positive difference indicates that the estimated reuse efficiency of the solution has not reached the preset index and further optimization is needed.

[0085] Subsequently, a difficulty assessment was conducted on the fragment consolidation of the initial migration and reuse scheme to obtain a difficulty value. The fragment consolidation difficulty assessment was primarily quantified based on the distribution of the number of free resource blocks not directly occupied by virtual machines in the region marking results and historical consolidation time data. The formula for calculating the difficulty value is: Difficulty Value = Number of Unoccupied Free Resource Blocks × Location Dispersion Coefficient × Historical Consolidation Time Coefficient. The location dispersion coefficient is calculated based on the physical location distribution of the free resource blocks; the more dispersed the distribution, the higher the coefficient value. The historical consolidation time coefficient is calculated based on the average time taken for fragment consolidation in the past; the longer the time, the higher the coefficient value. For example, if the target node has 3 remaining CPU free blocks and 2 remaining memory free blocks after virtual machine migration, distributed in different physical regions of the node, with a location dispersion coefficient of 1.5, and the historical average consolidation time is 1.2 times the node's average time, then the difficulty value = (3+2)×1.5×1.2=9. A higher difficulty value indicates a higher cost for subsequent consolidation of these fragments and a lower feasibility of the scheme.

[0086] Finally, the adaptation verification result is determined based on the adaptation difference, reuse difference, and difficulty value. During verification, preset thresholds for adaptation difference, reuse difference, and difficulty are established. The calculated three values ​​are compared with their corresponding thresholds to comprehensively determine whether the solution passes verification. For example, if the preset adaptation difference threshold is 1GB, the reuse difference threshold is 10%, and the difficulty threshold is 5, and the solution's adaptation difference is 2GB, reuse difference is 15%, and difficulty is 9, all exceeding the preset thresholds, then the solution fails verification and requires fragmentation consolidation optimization. If all three values ​​are below their corresponding thresholds, the solution passes verification and can be directly used as the basis for the final migration and reuse solution. The adaptation verification result not only clarifies whether the solution meets the preset requirements but also identifies specific problems in the solution through the three dimensions of values, such as insufficient adaptability, low reuse efficiency, or excessive consolidation difficulty. This provides direction for subsequent targeted optimization, ensuring that the final migration and reuse solution meets the virtual machine's resource requirements while effectively controlling fragmentation and improving overall resource utilization efficiency.

[0087] The initial migration and reuse scheme is subjected to fragment adaptation assessment to obtain the adaptation difference, which specifically includes the following steps:

[0088] Extract standard fragment adaptation benchmark data to obtain adaptation benchmark dataset;

[0089] Obtain the initial migration reuse scheme prediction adaptation data to obtain the prediction adaptation dataset;

[0090] The adaptation results are obtained by comparing the predicted adaptation dataset with the adaptation benchmark dataset.

[0091] The adaptation difference is obtained by calculating the difference between the adaptation results.

[0092] First, standard fragmentation adaptation benchmark data is extracted to obtain the adaptation benchmark dataset. This benchmark data is a reference standard derived from historical best-practice fragmentation and virtual machine migration cases in cloud data centers. It defines the ideal fragmentation matching state corresponding to different virtual machine resource requirements. The adaptation benchmark dataset contains multi-dimensional benchmark indicators, such as the required size of consecutive free resource blocks, fragmentation distribution concentration, and resource block continuity for virtual machines with specific CPU core counts, memory size, and disk capacity. For example, for a virtual machine requiring 4 CPU cores and 8GB of memory, the adaptation benchmark dataset records that, ideally, the target node should have a single consecutive 4-core CPU free block and a single consecutive 8GB memory free block, and the fragmentation distribution concentration should not be lower than a preset threshold. This benchmark data provides a unified evaluation standard for subsequent adaptation judgments, ensuring that the matching degree of different solutions can be fairly quantified and compared.

[0093] The predicted adaptation dataset is obtained by acquiring the predicted adaptation data of the initial migration and reuse scheme. Based on the initial migration and reuse scheme, the predicted adaptation dataset simulates and predicts the resource matching status after a virtual machine migrates to the target node. It reflects the actual fragmented resource usage of the virtual machine in the scheme. For example, in the initial scheme, a target node is matched for a virtual machine with 4 CPU cores and 8GB of memory. This node currently has two consecutive 2-core CPU free blocks, one consecutive 6GB memory free block, and one consecutive 2GB memory free block. The predicted adaptation dataset records that after the virtual machine migrates in, it will occupy these two 2-core CPU free blocks and the 6GB memory free block, as well as the distribution of the remaining 2GB memory free block, and the change in the concentration of node fragmentation after the virtual machine migrates in. The predicted adaptation dataset fully presents the actual matching status of the virtual machine and the target node's fragmented resources in the scheme, providing a direct basis for subsequent comparison with baseline data.

[0094] The adaptation results are obtained by comparing the predicted adaptation dataset with the adaptation benchmark dataset. The comparison process involves comparing each indicator in the predicted adaptation dataset with its corresponding indicator in the adaptation benchmark dataset one by one to determine the deviation between the actual matching state and the ideal state of the solution. For example, the benchmark data requires the virtual machine to occupy a single contiguous 4-core CPU free block and a single contiguous 8GB memory free block, while the predicted data shows the virtual machine occupying two 2-core CPU free blocks and one 6GB memory free block. Therefore, both the CPU resource block continuity indicator and the memory resource block size indicator deviate from the benchmark. Similarly, the benchmark data requires a fragmentation distribution concentration of no less than 70%, while the predicted data shows a fragmentation distribution concentration of 50% after the virtual machine migration, also indicating a deviation. The adaptation results record these deviations, clarifying which indicators the solution fails to meet the benchmark requirements, providing specific information for subsequent difference calculations.

[0095] The adaptation difference is calculated by performing a difference analysis on the adaptation results. The adaptation difference is a comprehensive quantitative result of the deviations in each aspect of the adaptation results, and the calculation formula is: Adaptation Difference = CPU Adaptation Deviation + Memory Adaptation Deviation + Disk Adaptation Deviation + Fragmentation Concentration Deviation. Where, CPU Adaptation Deviation = Baseline contiguous resource block size corresponding to the virtual machine's CPU requirements - Total actual contiguous resource block size used in the solution; the calculation method for memory and disk adaptation deviations is the same as for CPU; Fragmentation Concentration Deviation = Baseline Concentration - Predicted Concentration of the Solution. For example, the baseline data shows a CPU contiguous resource block size of 4 cores, while the actual contiguous resource block size used by the solution is 2 cores, resulting in a CPU adaptation deviation of 2 cores. The baseline memory contiguous resource block size is 8GB, while the solution actually uses 6GB, resulting in a memory adaptation deviation of 2GB. There is no disk deviation. The baseline fragmentation concentration is 70%, while the solution predicts 50%, resulting in a concentration deviation of 20%. The final adaptation difference is 2+2+0+20=24. The larger the difference, the greater the deviation between the actual fragmentation adaptation state of the solution and the ideal baseline, and the worse the adaptation. Targeted adjustments need to be made in subsequent fragmentation consolidation optimization to reduce deviation items and improve the fragmentation adaptation of the solution.

[0096] The resource reuse rate is analyzed based on the region labeling results to obtain the reuse difference, which includes the following steps:

[0097] Obtain standard reuse rate data to obtain a standard reuse dataset;

[0098] Obtain the predicted reuse rate data of the initial migration reuse scheme to obtain the predicted reuse dataset;

[0099] The reuse results are obtained by comparing the predicted reuse dataset with the standard reuse dataset;

[0100] The reuse results are quantified to obtain the reuse difference.

[0101] First, standard reuse rate data is obtained to create a standard reuse dataset. This standard reuse rate data is a reference benchmark extracted from historical best-practice operation and maintenance cases and preset efficiency targets in cloud data centers, defining the resource reuse level that virtual machine migration should achieve under different scenarios. The standard reuse dataset contains multi-dimensional indicators, such as the target node resource block utilization rate corresponding to different virtual machine resource requirements, the reuse efficiency benchmark value after fragment consolidation, and the expected reuse rate under different fragmentation distribution scenarios. For example, for a virtual machine requiring 4 CPU cores and 8GB of memory, the standard reuse dataset specifies that, ideally, when the total idle resources on the target node are 8 CPU cores and 16GB of memory, the virtual machine should occupy 4 consecutive CPU cores and 8GB of memory, the resource block utilization rate should reach over 75%, and the reuse efficiency benchmark value after fragment consolidation should be 80%, providing a unified reference standard for subsequent evaluation.

[0102] The predicted reuse dataset is obtained by acquiring the predicted reuse rate data of the initial migration reuse scheme. This dataset, combined with region labeling results, simulates and predicts the actual resource reuse after a virtual machine migrates to the target node, fully presenting the actual resource utilization status in the scheme. For example, in the initial scheme, a target node is matched for a virtual machine with 4 CPU cores and 8GB of memory. The region labeling results show that the node has two consecutive free 2-core CPU blocks, one free 6GB memory block, and one free 2GB memory block. After the virtual machine migrates in, it occupies these two 2-core CPU blocks and the 6GB memory block. The predicted reuse dataset records data on resource block utilization, fragmentation concentration changes, and the reuse potential of remaining free resource blocks in this scenario. For example, the current resource block utilization rate is 46%, far lower than the benchmark value in the standard reuse dataset, providing direct evidence for subsequent comparisons.

[0103] Then, the predicted reuse dataset is compared with the standard reuse dataset to obtain the reuse results. The comparison process involves comparing each indicator in the predicted dataset with the corresponding benchmark in the standard reuse dataset one by one to identify the deviation between the actual reuse state and the ideal state of the solution. For example, the standard reuse dataset specifies that the resource block utilization rate should be no less than 75%, while the predicted data shows only 46%; the standard specifies that the reuse efficiency after fragment consolidation should reach 80%, while the solution predicts that the reuse efficiency after consolidation is only 60%; simultaneously, the standard specifies that the proportion of remaining idle resource blocks that can be reused by subsequent virtual machines should be no less than 50%, while in the solution, the remaining resource blocks are highly dispersed, and the reusable proportion is only 20%. The reuse results will clearly record these deviations, indicating the gap between the solution and the standard requirements in terms of resource utilization efficiency.

[0104] Finally, the reuse results are quantified to obtain the reuse difference. The reuse difference is a comprehensive quantitative indicator of the deviations in the reuse results, and the calculation formula is: Reuse Difference = (Standard Resource Block Utilization Rate - Predicted Resource Block Utilization Rate) + (Standard Reuse Efficiency - Predicted Reuse Efficiency) + (Standard Reusable Ratio - Predicted Reusable Ratio). For example, if the standard resource block utilization rate is 75%, the prediction is 46%, and the deviation is 29%; the standard reuse efficiency is 80%, the prediction is 60%, and the deviation is 20%; the standard reusable ratio is 50%, the prediction is 20%, and the deviation is 30%. After comprehensive calculation, the reuse difference is 79%. The larger the difference, the greater the gap between the resource reuse efficiency of the scheme and the standard requirements. In subsequent optimization, it is necessary to focus on adjusting the resource matching method, improving the utilization rate of continuous resource blocks and the reusable potential of remaining fragments, and reducing the deviation from the standard.

[0105] The difficulty value is obtained by assessing the fragment integration difficulty of the initial migration and reuse scheme, which includes the following steps:

[0106] Obtain the fragment quantity data to get the fragment quantity value;

[0107] Obtain the dispersion value by acquiring fragment dispersion data;

[0108] The difficulty value is obtained by weighting the number of fragments and the dispersion value.

[0109] First, obtain the fragmentation count data to get the fragmentation count value. The fragmentation count data refers to the total number of remaining free resource blocks within the target node after the virtual machine is migrated, directly reflecting the severity of node fragmentation. For example, in the initial solution, a target node was matched with a virtual machine with a 4-core CPU and 8GB of memory. After the virtual machine migrates in, it occupies two 2-core CPU free blocks and one 6GB memory free block within the node, leaving one 2-core CPU free block and one 2GB memory free block remaining. In this case, the fragmentation count data is 2. The higher the fragmentation count, the more fragmented the free resources within the node are, resulting in more independent resource blocks to process during subsequent integration, and a higher basic integration cost.

[0110] Fragment dispersion data is obtained to determine the dispersion value. Fragment dispersion data measures the degree of concentration of remaining free resource blocks in the node's physical space. A higher value indicates a more dispersed distribution of resource blocks, making integration more difficult. Dispersion data calculation is typically based on fragment region labeling results, statistically analyzing the physical distance or continuity between free resource blocks. For example, if the remaining 2-core CPU free block is located in processor core group A, and the 2GB memory free block is located in memory channel B, their physical distance is relatively large, resulting in a higher dispersion value. If the two resource blocks are located in adjacent physical regions, the dispersion value will be lower. Dispersion data is also combined with the node's historical fragmentation records to statistically analyze integration time under similar distribution scenarios, further quantifying the impact of distribution on integration difficulty.

[0111] Finally, the difficulty value is calculated by weighting the fragmentation count and dispersion value. The formula for calculating the difficulty value is: Difficulty value = Fragmentation count × First weighting coefficient + Dispersion value × Second weighting coefficient. The first and second weighting coefficients can be preset based on data center operation and maintenance experience. Generally, the fragmentation count has a more significant direct impact on consolidation time; therefore, the first weighting coefficient can be set to 0.6, and the second weighting coefficient to 0.4. For example, if the fragmentation count is 2 and the dispersion is 0.8, the difficulty value is: 2 × 0.6 + 0.8 × 0.4 = 1.52. A higher difficulty value indicates higher operational and time costs for subsequent fragment consolidation, and lower feasibility. A lower difficulty value means the fragmentation state of the solution is easier to improve through subsequent operations, resulting in better overall migration stability and operational efficiency. By quantifying the difficulty value, operations personnel can clearly determine the differences in consolidation costs between different initial solutions, prioritizing solutions with lower difficulty values ​​for optimization, ensuring that virtual machine migration meets resource requirements while controlling the additional overhead of subsequent fragmentation cleanup.

[0112] Based on the adaptation verification results, the initial migration and reuse scheme is optimized by fragment integration to obtain the final migration and reuse scheme, which includes the following steps:

[0113] The direction and priority of fragment integration are determined based on the adaptation verification results.

[0114] Based on the integration direction and priority, resource allocation, fragment merging strategies, and migration order are adjusted to obtain the final migration and reuse plan.

[0115] First, the direction and priority of fragment consolidation are determined based on the adaptation verification results. The adaptation verification results include three core indicators: adaptation difference, reuse difference, and difficulty value. Each indicator corresponds to a different optimization direction. The adaptation difference reflects the mismatch between the virtual machine's resource requirements and the target node's continuous idle resource blocks. If this difference exceeds a preset threshold, it indicates insufficient resource continuity matching in the solution. The consolidation direction should focus on merging scattered idle resource blocks into continuous blocks, prioritizing the resolution of adaptation issues. The reuse difference reflects the gap between the actual resource reuse rate and the standard reuse rate. If this difference is high, it indicates that the solution's resource utilization efficiency has not met expectations. The consolidation direction needs to focus on optimizing the resource allocation method to improve the direct utilization rate of idle resources. The difficulty value reflects the cost of subsequent fragment consolidation. If the difficulty value is too high, it indicates that the fragment distribution of the current node is too scattered, and the consolidation cost exceeds the preset upper limit. In this case, the consolidation direction can be adjusted to replace the target node to avoid excessive subsequent maintenance costs. Integration priorities are ranked according to the severity of the three differences. For example, when the adaptation difference is far beyond the threshold, the resource continuity matching problem is addressed first, then the reuse efficiency is optimized, and finally the integration difficulty is assessed. This ensures that the optimization process first solves the core issues affecting the feasibility of the solution, and then gradually improves efficiency.

[0116] Next, the resource allocation fragmentation merging strategy and migration order are adjusted according to the integration direction and priority to obtain the final migration and reuse solution. Resource allocation adjustments must be combined with the integration direction. For example, to address the issue of excessively high compatibility discrepancies, the resource allocation position of virtual machines on the target node can be adjusted, prioritizing the use of contiguous free resource blocks to reduce the occupation of scattered fragments. If the target node does not have enough contiguous resource blocks, a fragmentation merging strategy can be used to merge adjacent free resource blocks within the node, creating contiguous resource space that meets the virtual machine's requirements. The fragmentation merging strategy needs to be formulated based on integration priority and difficulty. For example, fragments that are physically adjacent and have low integration difficulty should be merged first to reduce the time spent on merging operations and avoid creating new scattered fragments after merging. The migration order adjustment needs to be combined with the integration difficulty and the business priority of the virtual machines. For example, for virtual machines with low business priority, they can be migrated to the target node first, optimizing node resource distribution through fragmentation merging operations during the migration process without affecting core business operations. For virtual machines with high business priority, nodes that have already completed fragmentation merging should be selected first to ensure a stable migration process and high resource matching. For example, in the initial solution, the adaptation difference is 2GB, the reuse difference is 15%, and the difficulty is 7. The preset thresholds are an adaptation difference of 1GB, a reuse difference of 10%, and a difficulty of 5. The integration priorities are adaptation, reuse rate, and integration difficulty, respectively. During optimization, the memory resource allocation of the virtual machine is first adjusted from occupying 6GB of contiguous blocks and 2GB of scattered blocks to occupying 8GB of contiguous blocks after merging, thereby reducing the adaptation difference. Next, the CPU resource allocation is adjusted to prioritize the use of contiguous 4-core CPU blocks, thereby improving resource block utilization and narrowing the reuse difference. Finally, the fragmentation merging strategy is adjusted according to the difficulty value, merging only fragments that are physically adjacent, thereby reducing the integration difficulty. At the same time, the migration order is adjusted, migrating the virtual machine with lower business priority first to complete the fragmentation merging, and then migrating the target virtual machine, ultimately obtaining a final migration and reuse solution that simultaneously meets the requirements of adaptation, reuse rate, and cost.

[0117] The final migration reuse scheme is then validated to obtain migration optimization results, which includes the following steps:

[0118] The verification results are obtained by comparing the resource fragmentation rate after the final migration and reuse scheme is executed with the preset fragmentation threshold.

[0119] The verification results are compared with the preset fragmentation threshold.

[0120] If the verification result is less than or equal to the preset fragmentation threshold, then output the migration optimization qualified result;

[0121] If the verification result is greater than the preset judgment threshold, the fragment reorganization result will be output.

[0122] The verification result is obtained by comparing the resource fragmentation rate after the final migration and reuse plan is executed with the preset fragmentation threshold. Resource fragmentation rate is a core indicator for measuring the degree of fragmentation of node resources. The calculation formula is: Resource fragmentation rate = (Total size of non-contiguous free resource blocks in the node ÷ Total amount of free resources in the node) × 100%. After the final migration and reuse plan is executed, the distribution of free resources of the target node is recalculated, and the post-execution resource fragmentation rate is calculated. For example, after the virtual machine is migrated to the target node, the total amount of free resources is 8GB of memory and 4 CPU cores, of which the total size of non-contiguous free resource blocks is 2GB of memory and 1 CPU core. Substituting into the formula, the memory fragmentation rate = (2 ÷ 8) × 100% = 25%, and the CPU fragmentation rate = (1 ÷ 4) × 100% = 25%. The overall resource fragmentation rate is the average of the two, which is 25%. The preset fragmentation threshold is the upper limit of resource fragmentation set according to data center operation and maintenance standards, for example, set to 30%. The calculated post-execution resource fragmentation rate is compared with this threshold to obtain the verification result. This result directly reflects whether the fragmentation level of the node after the plan is executed is within an acceptable range.

[0123] The verification results are compared with a preset fragmentation threshold. The comparison process uses the preset fragmentation threshold as a benchmark to determine the magnitude of the verification result. For example, if the preset fragmentation threshold is 30%, and the resource fragmentation rate in the verification result is 25%, then the verification result is lower than the preset threshold. However, if the total size of the non-contiguous free resource blocks of the node after execution is 3GB of memory and 2 CPU cores, with a memory fragmentation rate of 37.5%, a CPU fragmentation rate of 50%, and an overall fragmentation rate of 43.75%, then the verification result is higher than the preset threshold. This step provides a direct basis for subsequent result determination, ensuring that the execution effect of each solution can be evaluated in a standardized manner.

[0124] If the verification result is less than or equal to the preset fragmentation threshold, a migration optimization qualified result is output. When the resource fragmentation rate after the solution execution does not exceed the preset threshold, it indicates that after the virtual machine is migrated to the target node, the node's idle resources still maintain a high degree of continuity, the fragmentation level is controlled within an acceptable range, no additional fragmentation operation is required, and the migration optimization process can be completed normally. For example, if the verification result is 25% and the preset threshold is 30%, the solution execution effect is determined to be satisfactory, a migration optimization qualified result is output, and the node can directly enter the normal operation state, providing a stable resource environment for subsequent virtual machine deployments.

[0125] If the verification result exceeds the preset threshold, the fragmentation re-compactment result is output. When the resource fragmentation rate after the solution execution exceeds the preset threshold, it indicates that the fragmentation problem of the node after the virtual machine migration has not been effectively alleviated, or has even worsened, requiring the subsequent fragmentation re-compactment process to be triggered. For example, if the verification result is 43.75%, exceeding the preset threshold of 30%, the fragmentation re-compactment result is output, prompting the operation and maintenance system to start the fragmentation operation. By merging discontinuous idle resource blocks, the resource fragmentation rate of the node is reduced, ensuring that the node resource status is restored to a level that meets the operation and maintenance standards, and preventing fragmentation problems from affecting the subsequent resource allocation and running performance of virtual machines.

[0126] 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; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for optimizing virtual machine migration in a cloud data center, characterized in that, The method includes the following steps: Acquire resource fragment data, virtual machine resource usage data, idle resource distribution data, and historical fragmentation records of physical nodes in the cloud data center to construct a resource fragment dataset; Fragment feature data is obtained by collecting fragment features from resource fragment datasets, and a set of target nodes is obtained by filtering high-reusability migration target nodes based on fragment feature data; By performing pattern statistics on historical fragment records, fragment pattern data is obtained, and a resource reuse optimization database is constructed based on the fragment pattern data. Input the resource requirements of the virtual machines to be migrated, the node fragmentation threshold, and the reuse efficiency indicators into the resource reuse optimization database to generate an initial migration and reuse plan. Fragment adaptation verification is performed on the initial migration and reuse scheme to obtain the adaptation verification result; Based on the adaptation verification results, the initial migration reuse scheme is integrated and optimized to obtain the final migration reuse scheme. The reuse effect of the final migration reuse scheme is verified to obtain the migration optimization results.

2. The cloud data center virtual machine migration optimization method according to claim 1, characterized in that, The resource fragment dataset includes CPU fragment data, memory fragment data, disk fragment data, free resource block size, fragment distribution location, and fragment defragmentation time data.

3. The cloud data center virtual machine migration optimization method according to claim 2, characterized in that, The target node set is obtained by filtering high-reusability migration target nodes based on fragment feature data, specifically including the following steps: A node reuse evaluation matrix is ​​constructed based on fragment feature data, and the reuse evaluation matrix is ​​classified into levels to obtain reuse level results; The physical node fragment data in the reuse level results are integrated to obtain the node fragment dataset; The resource reuse degree is calculated based on the node fragment dataset to obtain the reuse degree result; The target node set is formed based on the reuse rate results.

4. The cloud data center virtual machine migration optimization method according to claim 3, characterized in that, The initial migration and reuse scheme is subjected to fragment adaptation verification to obtain the adaptation verification result, which specifically includes the following steps: The initial migration and reuse scheme is marked with fragment region numbers to obtain region marking results, and the initial migration and reuse scheme is judged for fragment adaptation to obtain adaptation difference; The reuse rate is analyzed based on the regional labeling results to obtain the reuse difference; The difficulty value is obtained by assessing the difficulty of fragment integration of the initial migration and reuse scheme; The adaptation verification result is determined based on the adaptation difference, reuse difference, and difficulty value.

5. The cloud data center virtual machine migration optimization method according to claim 4, characterized in that, The initial migration and reuse scheme is subjected to fragment adaptation assessment to obtain the adaptation difference, which specifically includes the following steps: Extract standard fragment adaptation benchmark data to obtain adaptation benchmark dataset; Obtain the initial migration reuse scheme prediction adaptation data to obtain the prediction adaptation dataset; The adaptation results are obtained by comparing the predicted adaptation dataset with the adaptation benchmark dataset. The adaptation difference is obtained by calculating the difference between the adaptation results.

6. The cloud data center virtual machine migration optimization method according to claim 5, characterized in that, The resource reuse rate is analyzed based on the region labeling results to obtain the reuse difference, which includes the following steps: Obtain standard reuse rate data to obtain a standard reuse dataset; Obtain the predicted reuse rate data of the initial migration reuse scheme to obtain the predicted reuse dataset; The reuse results are obtained by comparing the predicted reuse dataset with the standard reuse dataset; The reuse results are quantified to obtain the reuse difference.

7. The cloud data center virtual machine migration optimization method according to claim 6, characterized in that, The difficulty value is obtained by assessing the fragment integration difficulty of the initial migration and reuse scheme, which includes the following steps: Obtain the fragment quantity data to get the fragment quantity value; Obtain the dispersion value by acquiring fragment dispersion data; The difficulty value is obtained by weighting the number of fragments and the dispersion value.

8. The cloud data center virtual machine migration optimization method according to claim 7, characterized in that, Based on the adaptation verification results, the initial migration and reuse scheme is optimized by fragment integration to obtain the final migration and reuse scheme, which includes the following steps: The direction and priority of fragment integration are determined based on the adaptation verification results. Based on the integration direction and priority, resource allocation, fragment merging strategies, and migration order are adjusted to obtain the final migration and reuse plan.

9. A cloud data center virtual machine migration optimization method according to claim 8, characterized in that, The final migration reuse scheme is then validated to obtain migration optimization results, which includes the following steps: The verification results are obtained by comparing the resource fragmentation rate after the final migration and reuse scheme is executed with the preset fragmentation threshold. The verification results are compared with the preset fragmentation threshold. If the verification result is less than or equal to the preset fragmentation threshold, then output the migration optimization qualified result; If the verification result is greater than the preset judgment threshold, the fragment reorganization result will be output.

Images