Virtual machine live migration optimization method and system based on multi-objective optimization

Abstract

The invention discloses a virtual machine live migration optimization method and system based on multi-objective optimization. A live migration optimization target is defined according to the virtual machine live migration influence factors, a multi-target optimization algorithm is initialized based on multiple factor items in the optimization target, and the multiple factor items comprise migration time, bandwidth occupation, resource consumption, migration success rate and performance overhead; the multi-objective optimization algorithm is used for outputting a real-time migration strategy according to a real-time memory access mode, predicting a dirty page rate change trend through trend fitting in a set sliding window, and dynamically adjusting the real-time migration strategy in combination with an access frequency; the synchronization frequency is adjusted in real time based on the dirty page rate change trend; and calculating the real-time load and the resource utilization rate of the target virtual machine during each incremental synchronization, evaluating the decompression capability of the target virtual machine, and adjusting the dirty page compression degree and the compression form. Through a multi-objective optimization algorithm and a real-time dynamic adjustment strategy, the fault tolerance and adaptability of the system under different load conditions are enhanced.

Description

Technical Field

[0001] This invention belongs to the field of virtual machine hot migration technology, specifically relating to a virtual machine hot migration optimization method and system based on multi-objective optimization. Background Technology

[0002] Virtual machine hot migration refers to the process of migrating a virtual machine from one physical server to another without shutting it down. During the migration, the virtual machine continues to run, without interrupting service or affecting the user experience. Hot migration typically involves steps such as memory state copying, disk state synchronization, CPU and network state migration, and switchover execution.

[0003] In existing technologies, optimization of hot migration efficiency largely focuses on the memory state copying step. For example, patent application CN115981780A identifies and skips the transmission of zero-page content, transmitting only its address information, allowing the destination to directly create zero-pages based on the address information, thereby reducing network bandwidth and computing resource consumption. Patent application CN116010034A uses dynamic monitoring and classification management to migrate memory blocks with low dirty page rates in a limited manner, reducing the number of repeated transmissions of hot memory blocks and improving migration efficiency. Patent application CN114443211A solves the hot migration convergence problem through dynamic rate limiting. It utilizes hardware PML functionality to track the dirty page rate of each VCPU in real time; if the total dirty page rate exceeds the migration transmission rate, it sorts the VCPUs by dirty page rate, selecting high-load target VCPUs; and uses Intel RDT / MBA technology to limit the memory bandwidth of the target VCPUs, specifically reducing their dirty page generation speed, rather than globally pausing the VCPUs. This method ensures migration convergence with minimal performance overhead, significantly improving the success rate of hot migration of high-load virtual machines.

[0004] Existing technologies have all improved the efficiency of virtual machine hot migration to some extent, but they share a common problem: existing patents address the optimization challenges of virtual machine hot migration efficiency from several dimensions, including data transmission redundancy, transmission scheduling strategies, and system resource balancing. However, hot migration efficiency often involves multiple factors such as computational complexity, bandwidth consumption, and performance overhead. Therefore, how to balance and optimize various overheads while ensuring migration efficiency, and guaranteeing the system's versatility and scalability, has become an urgent technical challenge. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a virtual machine hot migration optimization method and system based on multi-objective optimization. The hot migration optimization objective is defined according to the influencing factors of virtual machine hot migration. A multi-objective optimization algorithm is initialized based on multiple factors in the optimization objective, including migration time, bandwidth usage, resource consumption, migration success rate, and performance overhead. The multi-objective optimization algorithm outputs a real-time migration strategy based on real-time memory access patterns. Within a set sliding window, it predicts the dirty page rate change trend through trend fitting and dynamically adjusts the real-time migration strategy in conjunction with the access frequency. The synchronization frequency is adjusted in real-time based on the dirty page rate change trend. During each incremental synchronization, the real-time load and resource utilization of the target virtual machine are calculated, the decompression capability of the target virtual machine is evaluated, and the dirty page compression degree and compression method are adjusted.

[0006] The first aspect of this application discloses a virtual machine hot migration optimization method based on multi-objective optimization, which adopts the following technical solution: The hot migration optimization objective is defined based on the factors affecting virtual machine hot migration. A multi-objective optimization algorithm is initialized based on the multi-factor items in the optimization objective, including migration time, bandwidth usage, resource consumption, migration success rate, and performance overhead. The multi-objective optimization algorithm is used to output a real-time migration strategy based on the real-time memory access pattern. Within a set sliding window, the algorithm predicts the dirty page rate change trend through trend fitting and dynamically adjusts the real-time migration strategy in combination with the access frequency. The synchronization frequency is adjusted in real time based on the dirty page rate change trend; the real-time load and resource utilization of the target virtual machine are calculated during each incremental synchronization, the decompression capability of the target virtual machine is evaluated, and the dirty page compression degree and compression form are adjusted. During the migration process, the hot and cold sectors of the current virtual machine are continuously detected based on the real-time dirty page rate change trend, and the dirty page transfer priority is updated. The dirty page transfer volume, migration time, and resource parameters of the target virtual machine are monitored in real time. If an anomaly occurs during the migration, the migration strategy is replanned.

[0007] Furthermore, the multi-objective optimization algorithm is expressed in the following form: ; in, To comprehensively optimize the objective function, , , , and These represent migration time, bandwidth usage, resource consumption, migration success rate, and performance overhead, respectively. , , , and These are the weighting coefficients for migration time, bandwidth usage, resource consumption, migration success rate, and performance overhead, respectively; when initializing the multi-objective optimization algorithm, these are set based on historical data. , , , and The initial value.

[0008] Furthermore, the system monitors the resource usage of virtual machines in real time and performs periodic sampling to obtain load metrics; it also monitors the read and write behavior of each memory page in the virtual machine, records access patterns and access frequencies, and calculates the dirty page rate. Each set sliding window contains the following data: ,in Indicates setting a sliding window Data, express The dirty page rate at any given moment; Call the fitting algorithm to fit The dirty page rate variation is based on the set sliding window. Predicting from the data within, represented as ;in, For the predicted dirty page rate, and These are the regression coefficients obtained from the fitting process.

[0009] Furthermore, the migration strategy is adjusted based on the predicted dirty page rate, including: Based on the predicted dirty page rate Adjustment Dirty page migration frequency at any given time , represented as ;in, The migration frequency adjustment function is expressed as follows: ;in, Indicates the migration frequency before adjustment. for The dirty page rate at any given moment; Based on the predicted dirty page rate Adjustment Dirty page migration priority at any time , For memory page index; the The priority of a memory page is represented as ,in, Indicates the frequency of memory page access; and This is a balancing adjustment factor.

[0010] Furthermore, the process of adjusting the synchronization frequency in real time based on the dirty page rate change trend includes: The method for adjusting the synchronization frequency is expressed as follows: ;in, The adjusted synchronization frequency. Sensitivity coefficient for synchronization frequency adjustment; The sensitivity coefficient according to The settings include virtual machine load, memory access mode, bandwidth ratio, memory and migration latency tolerance at any given time.

[0011] Furthermore, the evaluation method for the target virtual machine's decompression capability is as follows: Calculate the real-time load of the target virtual machine and resource utilization ,based on and Calculate the decompression capability of the target virtual machine ; Defined as In the formula, For load error compensation; and To influence the weights.

[0012] Furthermore, the compression level adjustment method is as follows: ; This indicates the degree of compression, with a value range of [0,1], where 0 indicates no compression and 1 indicates maximum compression. This represents the maximum decompression capacity of the target virtual machine.

[0013] The second aspect of this application discloses a virtual machine hot migration optimization system based on multi-objective optimization, which implements the virtual machine hot migration optimization method as described in the first aspect of this application. The system includes: Initial definition module; Define hot migration optimization objectives based on factors affecting virtual machine hot migration, and initialize a multi-objective optimization algorithm based on multiple factors in the optimization objectives, including migration time, bandwidth usage, resource consumption, migration success rate, and performance overhead; The real-time migration strategy adjustment module uses the multi-objective optimization algorithm to output a real-time migration strategy based on the real-time virtual machine load and memory access patterns. Within a set sliding window, it predicts the dirty page rate change trend and change area through trend fitting and dynamically adjusts the real-time migration strategy. The decompression evaluation module adjusts the synchronization frequency in real time based on the dirty page rate change trend; calculates the real-time load and resource utilization of the target virtual machine during each incremental synchronization, evaluates the decompression capability of the target virtual machine, and adjusts the dirty page compression degree and compression form. The monitoring and feedback module detects hot and cold sectors of the current virtual machine based on the real-time dirty page rate change trend during the migration process and updates the dirty page transfer priority; it monitors the dirty page transfer volume, migration time, and resource parameters of the target virtual machine in real time, and re-plans the migration strategy when an anomaly occurs during migration.

[0014] The beneficial effects of this invention are that, compared with the prior art, 1. This application comprehensively considers multiple factors such as migration time, bandwidth usage, resource consumption, migration success rate, and performance overhead through a multi-objective optimization algorithm. By combining dirty page rate change trend prediction and dynamic adjustment of migration strategy, it can dynamically optimize migration frequency and dirty page priority during the migration process, minimize bandwidth usage and resource consumption, thereby improving the efficiency of virtual machine migration, reducing system load, ensuring optimal use of resources during migration, and reducing resource waste during hot migration.

[0015] 2. This application monitors the virtual machine's memory access patterns, load, and resource usage in real time, and adjusts the synchronization frequency and migration strategy based on the dirty page rate change trend. When an anomaly occurs during the migration process, the system can promptly replan the migration strategy to avoid migration failure due to sudden load or resource limitations. This dynamic adjustment mechanism effectively enhances the fault tolerance and adaptability of the virtual machine hot migration process, enabling the system to self-optimize under different load conditions and ensure successful migration. Attached Figure Description

[0016] Figure 1 A framework diagram for optimizing virtual machine hot migration methods. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of this invention. The embodiments described in this application are merely some embodiments of this invention, and not all embodiments. Based on the spirit of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of this invention.

[0018] Current optimizations of virtual machine hot migration methods typically focus on migration efficiency and success rate. However, finding an optimal balance among multiple dimensions such as algorithm performance, system overhead, and versatility is also a challenge that needs attention. Existing hot migration optimization strategies often emphasize a single objective without fully considering the combined impact of multiple objectives on hot migration performance or effectiveness. To address this issue, this application comprehensively considers multiple factors such as migration time, bandwidth usage, resource consumption, migration success rate, and performance overhead, and implements flexible and dynamic optimization adjustments among these objectives, thereby improving migration efficiency without sacrificing performance and stability.

[0019] As an embodiment of this application, a specific implementation of a virtual machine hot migration optimization method based on multi-objective optimization is disclosed. See [link to relevant documentation]. Figure 1 . Figure 1 A framework diagram for optimizing virtual machine hot migration methods.

[0020] As one implementation method of this embodiment, a hot migration optimization objective is defined based on the factors affecting virtual machine hot migration, and a multi-objective optimization algorithm is initialized based on the multi-factor items in the optimization objective.

[0021] In a further embodiment, the multi-factor items include migration time, bandwidth usage, resource consumption, migration success rate, and performance overhead. The multi-objective optimization algorithm is expressed as follows: ; in, To comprehensively optimize the objective function, , , , and These represent migration time, bandwidth usage, resource consumption, migration success rate, and performance overhead, respectively. , , , and These are the weighting coefficients for migration time, bandwidth usage, resource consumption, migration success rate, and performance overhead, respectively; when initializing the multi-objective optimization algorithm, these are set based on historical data. , , , and The initial value.

[0022] Among the multiple factors, migration time refers to the time required to complete the entire virtual machine hot migration. The goal is to minimize migration time and reduce the impact on the normal operation of the virtual machine. Bandwidth usage refers to the network bandwidth consumed during the migration process. The optimization goal is to reduce bandwidth usage and avoid excessive bandwidth consumption affecting the normal operation of other services. Resource consumption refers to the computing and storage resources consumed during the migration process. Reducing resource consumption can reduce the burden on the virtual machine host and target host resources. Migration success rate refers to the proportion of successful migration. The optimization goal is to improve the migration success rate and ensure the reliability of the migration. Performance overhead refers to the performance loss incurred during the migration process, including the proportion of resources such as CPU and memory. The optimization goal is to minimize performance overhead and ensure that the virtual machine maintains normal performance as much as possible during the migration.

[0023] In a further implementation, the purpose of initializing the multi-objective optimization algorithm is to generate a currently optimal initial migration strategy based on the current virtual machine's actual state, resource usage, and historical data.

[0024] The current state of a virtual machine includes, but is not limited to, current load, memory access mode, resource utilization (CPU, memory, etc.), dirty page rate, and network bandwidth. These multiple factors are calculated based on the current state of the virtual machine using standard methods, which will not be elaborated upon here.

[0025] In a further implementation, a solution algorithm is invoked to solve the multi-objective optimization function. Generate an initial migration strategy, with weights for each computational item. , , , and The initial values ​​can be designed based on historical data or previous experience. These weighting coefficients reflect the importance of each optimization objective in the current environment. The initial migration strategy is a decision set including but not limited to the following: migration start time, resource allocation of the target virtual machine, synchronization frequency, cold / hot zones, data compression format, bandwidth limits, priority order, etc.

[0026] In a specific, non-mandatory implementation, based on the successful experience of historical migration tasks, appropriate initial weights are designed by measuring the impact of each calculation item on the final success of the migration. During the uncertain phase of the migration, these weight coefficients can be adjusted based on real-time feedback to further optimize the migration strategy.

[0027] In a specific, non-mandatory implementation, the algorithm for solving the multi-objective optimization function can be a genetic algorithm or a particle swarm optimization algorithm.

[0028] As one implementation method of this embodiment, the migration task is executed according to the initial migration strategy. During the migration process, a multi-objective optimization algorithm is used to output a real-time migration strategy based on the real-time virtual machine load and memory access mode. Within a set sliding window, the trend of dirty page rate change is predicted by trend fitting, and the real-time migration strategy is dynamically adjusted.

[0029] In a further implementation, the read / write behavior of each memory page in the virtual machine is monitored, access patterns and access frequencies are recorded, and the dirty page rate is calculated; in this embodiment, it is based on time periods. The amount of dirty page data generated internally is used to calculate the dirty page rate. .

[0030] In a further embodiment, the size of the sliding window is set to... Each set sliding window contains the following data: ,in Indicates setting a sliding window Data, express The dirty page rate at any given moment; The dirty page rate change is fitted using a fitting algorithm based on a set sliding window. Predicting from the data within, represented as ;in, For the predicted dirty page rate, and These are the regression coefficients obtained from the fitting, representing the changing trend of the dirty page rate.

[0031] In a further implementation, adjusting the migration strategy based on the predicted dirty page rate includes: Based on the predicted dirty page rate Adjustment Dirty page migration frequency at any given time , represented as ;in, The migration frequency adjustment function is expressed as follows: ;in, Indicates the migration frequency before adjustment. for The dirty page rate at any given moment; Based on the predicted dirty page rate Adjustment Dirty page migration priority at any time , For memory page index; the The priority of a memory page is represented as ,in, Indicates the frequency of memory page access; and This is a balancing adjustment factor.

[0032] In one optional embodiment, the adjustment of dirty page migration frequency and priority can be considered not only in terms of memory access frequency and dirty page rate, but also in conjunction with the real-time load of the virtual machine. If the virtual machine load is high, memory pages with high access frequency and high dirty page rate are migrated first to avoid increasing the additional load on the virtual machine.

[0033] In a further implementation, after updating the dirty page rate and priority, the optimization algorithm is invoked to resolve the multi-objective optimization function, and the current migration strategy is updated. Specifically, the weight coefficients of the multi-factor calculation items are updated based on the current real-time state of the virtual machine.

[0034] As an example of this implementation, if excessive virtual machine load is detected, the impact of migration time and bandwidth usage on the migration strategy should increase, and therefore the corresponding weighting coefficients should be adjusted accordingly. and Improvements are made to ensure that the migration process does not result in excessive downtime or bandwidth strain.

[0035] As one implementation method of this embodiment, the synchronization frequency is adjusted in real time based on the dirty page rate change trend; during each incremental synchronization, the real-time load and resource utilization of the target virtual machine are calculated, the decompression capability of the target virtual machine is evaluated, and the dirty page compression degree and compression form are adjusted.

[0036] In a further embodiment, the process of adjusting the synchronization frequency in real time based on the dirty page rate change trend includes: The method for adjusting the synchronization frequency is expressed as follows: ; in, The adjusted synchronization frequency. Sensitivity coefficient for synchronization frequency adjustment; The sensitivity coefficient according to Virtual machine load at any time Memory access patterns (i.e., access frequency) Bandwidth ratio and migration delay tolerance General settings. Represented as: ; Unnormalized ; , , and These are the weighting coefficients for virtual machine load, memory access mode, bandwidth ratio, and migration latency tolerance, which can be set according to actual needs; This is the error effect term.

[0037] Migration bandwidth tolerance refers to the maximum migration latency that a system or application can tolerate during virtual machine hot migration. During hot migration, the virtual machine's running state, memory data, and execution environment need to be migrated from the source host to the target host, and this process inevitably introduces latency. Migration latency tolerance refers to the maximum time limit within which the virtual machine can run normally without service interruption. For applications with high real-time requirements, migration latency tolerance is lower, and the migration process must minimize latency to avoid impacting service continuity and user experience. Conversely, for applications with higher fault tolerance or less reliance on / requirement of real-time performance, migration latency tolerance is higher, and the sensitivity factor is correspondingly lower.

[0038] In a further implementation, the decompression capability of the target virtual machine is closely related to its real-time load and resource utilization. The evaluation method for the target virtual machine's decompression capability is as follows: Calculate the real-time load of the target virtual machine and resource utilization ,based on and Calculate the decompression capability of the target virtual machine ; Defined as In the formula, The load error compensation term represents the impact of factors other than computing resources, such as disk OI or bandwidth limitations. This embodiment uses the error compensation term to simplify the modeling of these factors. and To influence the weights.

[0039] The compression level adjustment method is as follows: ; This indicates the degree of compression, with a value range of [0,1], where 0 indicates no compression and 1 indicates maximum compression. This represents the maximum decompression capacity of the target virtual machine.

[0040] In a further embodiment, if A value less than 1 indicates that the target virtual machine can handle the data, and strong compression is not necessary. For example... At this point, a compression level of 40% is used. A faster compression algorithm can be selected to ensure overall efficiency.

[0041] like A value greater than or equal to 1 indicates that dirty pages are being generated too quickly, and even the target virtual machine cannot handle the full decompression. In this case, the highest level of compression, namely 100% compression, is forcibly applied.

[0042] As one implementation method of this embodiment, during the migration process, the hot and cold sectors of the current virtual machine are detected based on the real-time dirty page rate change trend, and the dirty page transfer priority is updated; the dirty page transfer volume, migration time, and resource parameters of the target virtual machine are monitored in real time, and the migration strategy is replanned when an anomaly occurs during migration.

[0043] In a further real-time approach, during virtual machine migration, the access status of memory regions is continuously assessed by monitoring the dirty page rate of the virtual machine in real time. The dirty page rate refers to the rate at which modified data pages in memory change, reflecting the activity level of the virtual machine's memory.

[0044] Optionally, dirty pages in hot zones have high priority and should be transferred as early as possible to reduce the possibility of data loss or duplicate transfers during the migration process. Dirty pages in cold zones have low priority and can be migrated later or incrementally synchronized in the later stages of the migration. During the migration process, in addition to real-time monitoring of the dirty page rate, it is also necessary to monitor the dirty page transfer volume, migration time, and virtual machine resource utilization. In particular, metrics such as migration time and bandwidth usage will significantly impact adjustments to the migration strategy. Key metrics for real-time monitoring include: Dirty page transfers: Monitor the number of dirty pages transferred during the migration process to assess the migration progress.

[0045] Migration Time: Monitor the actual migration time to determine if it meets the expected migration time window.

[0046] Resource usage: Real-time tracking of virtual machine resource usage such as CPU, memory, and network bandwidth to prevent resource bottlenecks during migration.

[0047] When anomalies occur during the migration process (such as excessively long migration times, high bandwidth usage, slow virtual machine response, etc.), the migration strategy needs to be replanned. This anomaly detection is achieved through the following methods: Migration timeout: If the migration exceeds the preset maximum tolerance time, a replanning will be triggered.

[0048] Bandwidth overload: If bandwidth usage exceeds the set threshold, the migration frequency should be reduced or the migration should be delayed.

[0049] Resource bottleneck: If the CPU or memory resource utilization of the virtual machine exceeds the set threshold, the dirty page synchronization frequency should be reduced or the synchronization priority should be adjusted.

[0050] The process of re-planning a migration strategy includes the following steps: Dynamically adjust the synchronization frequency: The synchronization frequency of the migration is dynamically adjusted based on the current dirty page rate, migration progress, and resource load.

[0051] Update migration priorities: Reassess hot and cold zones, and update dirty page transfer priorities.

[0052] Migration path replanning: Based on the current migration progress and the status of the target virtual machine, re-invoke the multi-objective optimization solution, adjust the migration strategy and transmission order, and optimize migration efficiency.

[0053] As an embodiment of this application, a virtual machine hot migration optimization system based on multi-objective optimization is provided, employing a specific implementation method of the virtual machine hot migration optimization method described in the above-described method embodiments. The system includes the following modules: Initial definition module; Define hot migration optimization objectives based on factors affecting virtual machine hot migration, and initialize a multi-objective optimization algorithm based on multiple factors in the optimization objectives, including migration time, bandwidth usage, resource consumption, migration success rate, and performance overhead; The real-time migration strategy adjustment module uses the multi-objective optimization algorithm to output a real-time migration strategy based on the real-time virtual machine load and memory access patterns. Within a set sliding window, it predicts the dirty page rate change trend and change area through trend fitting and dynamically adjusts the real-time migration strategy. The decompression evaluation module adjusts the synchronization frequency in real time based on the dirty page rate change trend; calculates the real-time load and resource utilization of the target virtual machine during each incremental synchronization, evaluates the decompression capability of the target virtual machine, and adjusts the dirty page compression degree and compression form. The monitoring and feedback module continuously detects the hot and cold sectors of the current virtual machine based on the real-time dirty page rate change trend during the migration process, updates the dirty page transfer priority, monitors the dirty page transfer volume, migration time, and resource parameters of the target virtual machine in real time, and re-plans the migration strategy when an anomaly occurs during migration.

[0054] As an embodiment of this application, an electronic device is disclosed, including a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the computer program, when loaded onto the processor, implements the virtual machine hot migration optimization method according to the above method embodiment.

[0055] As an embodiment of this application, a computer-readable storage medium is disclosed, which stores a computer program that, when executed by a processor, implements the virtual machine hot migration optimization method according to the above method embodiments.

[0056] 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 it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A virtual machine hot migration optimization method based on multi-objective optimization, characterized in that, include: The hot migration optimization objective is defined based on the factors affecting virtual machine hot migration. A multi-objective optimization algorithm is initialized based on the multi-factor items in the optimization objective, including migration time, bandwidth usage, resource consumption, migration success rate, and performance overhead. The multi-objective optimization algorithm is used to output a real-time migration strategy based on the real-time memory access pattern. Within a set sliding window, the algorithm predicts the dirty page rate change trend through trend fitting and dynamically adjusts the real-time migration strategy in combination with the access frequency. The synchronization frequency is adjusted in real time based on the trend of the dirty page rate change. During each incremental synchronization, the real-time load and resource utilization of the target virtual machine are calculated, the decompression capability of the target virtual machine is evaluated, and the dirty page compression degree and compression method are adjusted. During the migration process, the hot and cold sectors of the current virtual machine are continuously detected based on the real-time dirty page rate change trend, and the dirty page transfer priority is updated accordingly. Monitor dirty page transfer volume, migration time, and resource parameters of the target virtual machine in real time. If an anomaly occurs during migration, replan the migration strategy.

2. The virtual machine hot migration optimization method based on multi-objective optimization according to claim 1, characterized in that, The multi-objective optimization algorithm is expressed in the following form: ； in, To comprehensively optimize the objective function, , , , and These represent migration time, bandwidth usage, resource consumption, migration success rate, and performance overhead, respectively. , , , and These are the weighting coefficients for migration time, bandwidth usage, resource consumption, migration success rate, and performance overhead, respectively; when initializing the multi-objective optimization algorithm, these are set based on historical data. , , , and The initial value.

3. The virtual machine hot migration optimization method based on multi-objective optimization according to claim 1, characterized in that, Real-time monitoring of virtual machine resource usage and periodic sampling to obtain load metrics; monitoring read and write behavior of each memory page in the virtual machine, recording access patterns and access frequency, and calculating dirty page rate; Each set sliding window contains the following data: ,in Indicates setting a sliding window Data, express The dirty page rate at any given moment; Call the fitting algorithm to fit The dirty page rate variation is based on the set sliding window. Predicting from the data within, represented as ;in, For the predicted dirty page rate, and These are the regression coefficients obtained from the fitting process.

4. The virtual machine hot migration optimization method based on multi-objective optimization according to claim 3, characterized in that, Adjust migration strategies based on the predicted dirty page rate, including: Based on the predicted dirty page rate Adjustment Dirty page migration frequency at any given time , represented as ;in, The migration frequency adjustment function is expressed as follows: ;in, Indicates the migration frequency before adjustment. for The dirty page rate at any given moment; Based on the predicted dirty page rate Adjustment Dirty page migration priority at any time , For memory page index; the The priority of a memory page is represented as ,in, Indicates the frequency of memory page access; and This is a balancing adjustment factor.

5. The virtual machine hot migration optimization method based on multi-objective optimization according to claim 1, characterized in that, The process of adjusting the synchronization frequency in real time based on the dirty page rate change trend includes: The method for adjusting the synchronization frequency is expressed as follows: ;in, The adjusted synchronization frequency. Sensitivity coefficient for synchronization frequency adjustment; The sensitivity coefficient according to The settings include virtual machine load, memory access mode, bandwidth ratio, memory and migration latency tolerance at any given time.

6. The virtual machine hot migration optimization method based on multi-objective optimization according to claim 1, characterized in that, The evaluation method for the target virtual machine's decompression capability is as follows: Calculate the real-time load of the target virtual machine and resource utilization ,based on and Calculate the decompression capability of the target virtual machine ; Defined as In the formula, For load error compensation; and To influence the weights.

7. The virtual machine hot migration optimization method based on multi-objective optimization according to claim 1, characterized in that, The compression level adjustment method is as follows: ; This indicates the degree of compression, with a value range of [0,1], where 0 indicates no compression and 1 indicates maximum compression. This represents the maximum decompression capacity of the target virtual machine.

8. A virtual machine hot migration optimization system based on multi-objective optimization, running the virtual machine hot migration optimization method as described in any one of claims 1-7, characterized in that, The system includes: Initial definition module; Define hot migration optimization objectives based on factors affecting virtual machine hot migration, and initialize a multi-objective optimization algorithm based on multiple factors in the optimization objectives, including migration time, bandwidth usage, resource consumption, migration success rate, and performance overhead; The real-time migration strategy adjustment module uses the multi-objective optimization algorithm to output a real-time migration strategy based on the real-time virtual machine load and memory access patterns. Within a set sliding window, it predicts the dirty page rate change trend and change area through trend fitting and dynamically adjusts the real-time migration strategy. The decompression evaluation module adjusts the synchronization frequency in real time based on the dirty page rate change trend; calculates the real-time load and resource utilization of the target virtual machine during each incremental synchronization, evaluates the decompression capability of the target virtual machine, and adjusts the dirty page compression degree and compression form. The monitoring and feedback module detects hot and cold sectors of the current virtual machine based on the real-time dirty page rate change trend during the migration process and updates the dirty page transfer priority; it monitors the dirty page transfer volume, migration time, and resource parameters of the target virtual machine in real time, and re-plans the migration strategy when an anomaly occurs during migration.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the computer program is loaded into the processor, it implements the virtual machine hot migration optimization method according to any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the virtual machine hot migration optimization method according to any one of claims 1-7.

Images