Server release method

Abstract

The application discloses a server release method, relates to the technical field of computers, determines a network quality index through a network state parameter of a physical host, determines active physical hosts and non-active physical hosts in a physical host cluster according to the network quality index, determines a virtual machine migration strategy through a multi-target optimization model and the network quality index, and migrates virtual machines out of the active physical hosts to shut down the active physical hosts. The physical hosts in the physical host cluster are screened through the network quality index, virtual machine migration is only performed on active physical hosts with good network quality, the reliability in the virtual machine migration process can be ensured, the multi-target optimization model is used, energy consumption, user request response time and virtual machine migration success rate are fully considered, and the balance between service quality and energy consumption optimization can be achieved.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a server release method. Background Technology

[0002] With the rapid development of IoT technology, edge nodes need to handle dynamically changing user requests while balancing resource utilization and energy consumption. Virtual machine migration technology, which releases servers as a core means of edge computing resource management, faces multiple challenges, including node mobility, network fluctuations, and hardware heterogeneity.

[0003] In related technologies, server release aims to reduce the number of physical hosts by migrating virtual machines off the physical hosts and then shutting down the physical hosts. However, the above solutions only take energy consumption reduction as an optimization goal, resulting in an imbalance between service quality and energy consumption optimization. Summary of the Invention

[0004] This application provides a server release method to at least address the problem of imbalance between service quality and energy consumption optimization in related technologies.

[0005] This application provides a server release method, including:

[0006] Obtain network status parameters of at least two physical hosts in the physical machine cluster, and determine the network quality index based on the network status parameters. The network quality index is used to indicate the active and inactive physical hosts in the physical machine cluster.

[0007] A multi-objective optimization model is obtained, and a virtual machine migration strategy is generated based on the multi-objective optimization model and the network quality index. The optimization objectives of the multi-objective optimization model include at least a first optimization objective and a second optimization objective. The first optimization objective is used to minimize the number of active physical hosts in the physical machine cluster, and the second optimization objective is used to minimize the response time of user requests to active physical hosts in the physical machine cluster, or to maximize the success rate of virtual machine migration between physical hosts in the physical machine cluster.

[0008] According to the virtual machine migration policy, at least one virtual machine within the target active physical host is migrated out in order to shut down the target active physical host.

[0009] This application also provides a server release device, comprising:

[0010] The acquisition module is used to acquire network status parameters of at least two physical hosts in the physical machine cluster, and determine the network quality index based on the network status parameters. The network quality index is used to indicate the active and inactive physical hosts in the physical machine cluster.

[0011] The first processing module is used to obtain a multi-objective optimization model and generate a virtual machine migration strategy based on the multi-objective optimization model and the network quality index. The optimization objectives of the multi-objective optimization model include at least a first optimization objective and a second optimization objective. The first optimization objective is used to minimize the number of active physical hosts in the physical machine cluster, and the second optimization objective is used to minimize the response time of user requests to active physical hosts in the physical machine cluster, or to maximize the success rate of virtual machine migration between physical hosts in the physical machine cluster.

[0012] The second processing module is used to migrate at least one virtual machine from the target active physical host according to the virtual machine migration policy, so as to shut down the target active physical host.

[0013] This application also provides an electronic device, including: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of any of the above-described server release methods.

[0014] This application also provides a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, it implements the steps of any of the above-described server release methods.

[0015] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of any of the above-described server release methods.

[0016] This application determines a network quality index based on the network status parameters of physical hosts, and then identifies active and inactive physical hosts within the physical machine cluster based on this index. A multi-objective optimization model and the network quality index are then used to determine a virtual machine migration strategy, migrating virtual machines from active physical hosts to shut down those hosts. By using the network quality index to filter physical hosts within the cluster, virtual machine migration is performed only on active physical hosts with good network quality, ensuring reliability during the migration process. The multi-objective optimization model fully considers energy consumption, user request response time, and virtual machine migration success rate, achieving a balance between service quality and energy efficiency. Attached Figure Description

[0017] To more clearly illustrate the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A schematic diagram of a hardware architecture provided for an embodiment of this application;

[0019] Figure 2 A flowchart illustrating the server release method provided in this application embodiment. Figure 1 ;

[0020] Figure 3 A flowchart illustrating the server release method provided in this application embodiment. Figure 2 ;

[0021] Figure 4 for Figure 3 A flowchart illustrating the specific implementation of step S303 in the illustrated embodiment;

[0022] Figure 5 for Figure 3 A flowchart illustrating the specific implementation of step S304 in the illustrated embodiment;

[0023] Figure 6 for Figure 3 A flowchart illustrating the specific implementation of step S305 in the illustrated embodiment;

[0024] Figure 7 for Figure 3 A flowchart illustrating the specific implementation of step S306 in the illustrated embodiment;

[0025] Figure 8 This is a schematic diagram of the server release device provided in an embodiment of this application;

[0026] Figure 9 A schematic diagram of the structure of the electronic device provided in this application. Detailed Implementation

[0027] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of this application.

[0028] It should be noted that, in the description of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. The terms "first," "second," etc., in this application are used to distinguish similar objects and are not used to describe a specific order or sequence.

[0029] With the rapid development of technologies such as the Internet of Things and autonomous driving, edge computing, with its significant advantages of low latency and high real-time performance, has become a key technology supporting various real-time applications. Edge nodes need to handle dynamically changing user requests while balancing resource utilization and energy consumption.

[0030] Server release refers to migrating virtual machines (VMs) on physical hosts with low resource utilization within a physical machine cluster to other physical hosts and then shutting down those physical hosts. Specifically, some physical hosts in a cluster may experience prolonged periods of low load, such as CPU utilization below 30% for several consecutive periods or failure to process user requests. Continuing to run such a physical host would waste resources, but directly shutting it down could disrupt the services it is responsible for. Therefore, it is necessary to migrate the VMs on that physical host to other physical hosts to shut down the host. Server release, as a core method of edge computing resource management, faces multiple challenges, including node mobility, network fluctuations, and hardware heterogeneity.

[0031] In related technologies, server release methods mostly focus on single-objective optimization, that is, reducing the number of physical hosts as the primary goal. This is achieved by predicting the resource requirements for the next cycle, determining the number of physical hosts that need to be shut down, and then migrating virtual machines from some physical hosts to other physical hosts and shutting them down to reduce energy consumption. However, in practical applications, the state of edge networks changes rapidly, user requests are sudden, and edge node hardware configurations vary. Related technologies suffer from the following problems:

[0032] 1. Focusing solely on reducing the number of physical hosts without considering user request response time and service reliability results in an imbalance between service quality and energy consumption optimization.

[0033] 2. It does not take into account the bandwidth fluctuations and latency changes caused by the mobility of edge nodes. The virtual machine migration decision relies on static resource prediction, which can easily lead to transmission timeouts or service interruptions.

[0034] 3. The prediction accuracy for the resources required in the next cycle is limited, and the adaptability to sudden requests (such as peak traffic) is poor. There is a problem of resource waste or service unavailability due to prediction deviation.

[0035] 4. Migration strategies were not optimized for differences in CPU architecture and memory capacity of edge nodes, resulting in a situation where high-configuration nodes have idle resources while low-configuration nodes are overloaded.

[0036] 5. The lack of fault tolerance mechanisms during the migration process may lead to data loss or request failure.

[0037] Based on this, this application proposes a server release method. It determines the network quality index using the network status parameters of physical hosts, and then identifies active and inactive physical hosts in the physical machine cluster based on the network quality index. A virtual machine migration strategy is determined using a multi-objective optimization model and the network quality index, migrating virtual machines from active physical hosts to shut down those hosts. By filtering physical hosts in the physical machine cluster using the network quality index, virtual machine migration is only performed on active physical hosts with good network quality, ensuring reliability during the migration process. The multi-objective optimization model fully considers energy consumption, user request response time, and virtual machine migration success rate, achieving a balance between service quality and energy consumption optimization.

[0038] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0039] This section describes the specific application environment architecture or hardware architecture that the server release method depends on. (References) Figure 1 , Figure 1 This is a schematic diagram of a hardware architecture provided for an embodiment of this application. In the field of edge computing, a physical machine cluster includes multiple physical hosts, such as physical host 1 to physical host M. These physical hosts can communicate with each other to jointly provide computing services to nearby terminal devices. In practical applications, some physical machines in the physical machine cluster can be shut down based on the number of requests initiated by the terminal devices and the required computing resources to reduce system energy consumption.

[0040] Figure 2 A flowchart illustrating the server release method provided in this application embodiment. Figure 1 ,like Figure 2 As shown, an embodiment of this application provides a server release method, which is described in detail below:

[0041] S201. Obtain the network status parameters of at least two physical hosts in the physical machine cluster, and determine the network quality index based on the network status parameters.

[0042] The executing entity of this application embodiment can be Figure 1 Any of the physical hosts shown can also be used with Figure 1 The computing device that enables communication between the physical hosts in the system.

[0043] A physical machine cluster is a resource collection consisting of multiple physical hosts with computing, storage, and networking capabilities in an edge computing environment. It is used to host virtual machines (VMs) and handle user requests. Physical hosts within a physical machine cluster can exhibit hardware heterogeneity, such as different CPU architectures or varying memory capacities. A physical host is a physical server with independent hardware resources within the physical machine cluster.

[0044] Network status parameters are used to evaluate the real-time network status of physical hosts. These parameters include network bandwidth, end-to-end latency, physical host movement speed, and signal strength. The Network Quality Index (NQI) is a comprehensive indicator obtained by quantifying network status parameters. It is used to classify physical hosts in a physical machine cluster, such as categorizing them into hosts with good network quality and those with poor network quality.

[0045] For example, network status parameters of at least two physical hosts in a physical machine cluster can be obtained in the following way, and the network quality index can be determined based on the network status parameters: the network status of the physical hosts is collected in real time according to a preset collection frequency to obtain network status parameters, which include at least one of the following: network bandwidth, end-to-end latency, physical host movement speed, and signal strength; the network status parameters are weighted and summed to obtain the network quality index, and the network quality index of the active physical host is better than that of the inactive physical host.

[0046] Network status parameters of physical hosts can be obtained by deploying monitoring agents on each physical host. Specifically, the monitoring agents collect parameters such as network bandwidth, end-to-end latency, physical host movement speed, and signal strength in real time at a preset collection frequency, such as 100ms / time.

[0047] The network quality index can be determined using the following formula:

[0048]

[0049] Among them, latency, bandwidth, and mobility speed are network status parameters acquired in real time. The weights are for latency, bandwidth, and movement speed, respectively, and their sum is 1. It can be set according to actual application needs; for example, it can be increased when users have high requirements for network latency. The value can be increased when high bandwidth requirements are needed. The value can be increased when the physical host is in a rapid movement state. The values ​​are as follows; maximum latency, maximum bandwidth, and maximum mobility are the design limits for edge networks.

[0050] Furthermore, physical hosts can be categorized into active and inactive physical hosts based on NQI. Specifically, NQI is divided into 5 levels (1-5) according to a preset threshold, with level 1 representing the best network condition and level 5 representing the worst. When NQI ≤ 3, the physical host is classified as an active physical host; when NQI > 3, the physical host is classified as an inactive physical host. Active physical hosts indicate good network conditions and can be migrated for virtual machines, while inactive physical hosts indicate poor network conditions, and migrating virtual machines for them may lead to transmission timeouts or service interruptions. Therefore, virtual machine migration is not performed on inactive physical hosts; that is, virtual machine migration is performed between active physical hosts, moving a virtual machine from one active physical host to another. Furthermore, network optimization can be performed on inactive physical hosts, and once their NQI ≤ 3, virtual machine migration can be resumed.

[0051] In the above method, physical hosts are divided into active and inactive physical hosts by network quality index. This ensures that virtual machine migration is carried out only on active physical hosts, avoiding transmission timeouts or service interruptions caused by migrating inactive physical hosts with poor network quality, and improving service reliability.

[0052] S202. Obtain the multi-objective optimization model, and generate a virtual machine migration strategy based on the multi-objective optimization model and the network quality index.

[0053] A multi-objective optimization model is a pre-built model used to determine virtual machine migration strategies. It can include multiple optimization objectives and constraints. Specifically, the optimization objectives of the multi-objective optimization model include at least a first optimization objective and a second optimization objective. The first optimization objective minimizes the number of active physical hosts within the physical machine cluster, and the second optimization objective minimizes the response time for user requests to active physical hosts within the physical machine cluster, or maximizes the success rate of virtual machine migration between physical hosts within the physical machine cluster. Constraints are the conditions that must be met to achieve the optimization objectives, such as resource utilization constraints and virtual machine migration time constraints.

[0054] In one possible implementation, the optimization objectives of the multi-objective optimization model include at least a first optimization objective, a second optimization objective, and a third optimization objective, and the model parameters corresponding to the optimization objectives satisfy at least one of the constraints of resource constraints, migration time constraints, and compatibility constraints.

[0055] The first optimization objective is to minimize energy consumption based on the number of active physical hosts and the total number of physical hosts; the second optimization objective is to minimize the response time of user requests based on the processing power of active physical hosts and the distance between the user and the active physical hosts; the third optimization objective is to maximize the success rate of virtual machine migration between physical hosts within the physical machine cluster based on historical migration data, which includes the number of migration failures, data loss, and total data volume during virtual machine migration in the previous period.

[0056] The optimization objective of a multi-objective optimization model can be determined by the following formula:

[0057]

[0058] in, The formula in question is the objective function corresponding to the first optimization objective. The formula in question is the objective function corresponding to the second optimization objective. The formula is the objective function corresponding to the third optimization objective, where n is the number of user requests. This represents the geographical distance (e.g., 100 meters) between the i-th user and the physical host where the requested VM is located. The overall performance of a physical host, including CPU core count and memory, can be represented by benchmark scores. A smaller value for the objective function corresponding to the first optimization objective indicates lower energy consumption; a smaller value for the objective function corresponding to the second optimization objective indicates shorter response times for user requests; and a larger value for the objective function corresponding to the third optimization objective indicates a higher success rate for virtual machine migration, i.e., higher reliability.

[0059] Resource constraints are used to limit the resource utilization of the active physical host after virtual machine migration to not exceed the first resource utilization threshold; migration time constraints are used to limit the migration time during virtual machine migration to not exceed the response time threshold; compatibility constraints are used to ensure that the operating system and hardware architecture of the virtual machine and the active physical host are compatible.

[0060] The first resource utilization threshold is, for example, 90%. This means that after a virtual machine migration, the resource utilization of the active physical host receiving the migrated virtual machine should not exceed 90% to ensure that the active physical host can operate normally. Resource utilization includes CPU resource utilization and memory resource utilization, that is, CPU resource utilization should not exceed 90% and memory resource utilization should not exceed 90%.

[0061] The response time threshold is a pre-set threshold for the duration of responding to user requests. The time taken during virtual machine migration should not exceed the response time threshold to ensure user experience.

[0062] Compatibility constraints are used to ensure that virtual machines match the operating system and hardware architecture of active physical hosts. Physical hosts within a physical machine cluster can have hardware heterogeneity, and virtual machines can have different performance and parameter configurations in order to match the physical hosts. Therefore, when migrating virtual machines, it is necessary to ensure that when a virtual machine is migrated to an active physical host, the active physical host should match the operating system and hardware architecture of the virtual machine.

[0063] A virtual machine migration policy is a scheme used to instruct the migration of virtual machines between active physical hosts, such as migrating a virtual machine from active physical host a to active physical host b, or migrating a virtual machine from active physical host c to active physical host b.

[0064] The virtual machine migration strategy is determined based on a multi-objective optimization model and a network quality index. That is, the virtual machine migration strategy should be a migration scheme that minimizes the number of active physical hosts in the physical machine cluster, minimizes the response time of user requests, and maximizes the success rate of virtual machine migration under the current network quality and constraints.

[0065] The above method, by using a multi-objective optimization model that simultaneously considers energy consumption, user request response time, and virtual machine migration success rate, and sets constraints, effectively overcomes the limitation of single-objective optimization that only focuses on a single indicator. It achieves a multi-dimensional balance between energy consumption control, service efficiency, and migration reliability in edge computing server resource management. At the same time, it provides a clear optimization direction and quantitative evaluation standard for the generation of subsequent virtual machine migration strategies.

[0066] S203. According to the virtual machine migration policy, migrate at least one virtual machine from the target active physical host to shut down the target active physical host.

[0067] The target active physical host is the active physical host from which virtual machines need to be migrated. After the virtual machines are migrated, the target active physical host will no longer process user requests. To reduce system power consumption, the target active physical host can be shut down, for example, by powering off or entering a low-power mode. Furthermore, data such as the shutdown time of the target active physical host and the resource savings corresponding to the shutdown operation can be recorded.

[0068] The server release method provided in this application determines the network quality index through the network status parameters of physical hosts, and identifies active and inactive physical hosts in the physical machine cluster based on the network quality index. A virtual machine migration strategy is determined through a multi-objective optimization model and the network quality index, migrating virtual machines from active physical hosts to shut down those hosts. By filtering physical hosts in the physical machine cluster using the network quality index, virtual machine migration is performed only on active physical hosts with good network quality, ensuring reliability during the virtual machine migration process. The multi-objective optimization model fully considers energy consumption, user request response time, and virtual machine migration success rate, achieving a balance between service quality and energy consumption optimization.

[0069] Figure 3 A flowchart illustrating the server release method provided in this application embodiment. Figure 2 ,like Figure 3 As shown, an embodiment of this application provides a server release method, which is described in detail below:

[0070] S301. Obtain the network status parameters of at least two physical hosts in the physical machine cluster, and determine the network quality index based on the network status parameters.

[0071] S302, Obtain historical request data.

[0072] Historical request data refers to user request data over a past period, such as user request data from the past 7 days. Historical request data includes request type, resource requirements, and request arrival time. Historical request data can be stored on the current computing device or in cloud storage and can be retrieved through data interfaces or direct access.

[0073] S303. Based on historical request data and network quality index, predict the resource demand for the next time period to obtain the target resource demand.

[0074] The time period is a pre-set duration, such as 5 minutes. The target resource requirement includes at least two target user requests, along with the corresponding resource requirements and request priorities. This target resource requirement is used to reserve active physical hosts that will not be shut down, to handle potential user requests in the next time period. The target resource requirement can be predicted using a trained predictive model.

[0075] For example, such as Figure 4 As shown, step S303 can be achieved by the following steps:

[0076] S3031. Obtain the target prediction model.

[0077] The target prediction model is a trained Long Short-Term Memory (LSTM) network model.

[0078] In one possible implementation, the target prediction model has an input layer consisting of 10 feature layers, 3 hidden layers with 32 neurons per layer, and an output layer containing the predicted resource demand for the next cycle.

[0079] Optionally, the LSTM model can be trained with training data to obtain the target prediction model. Specifically, the training data can be historical request data over a longer period. By using the Adam optimizer, the learning rate can be adaptively adjusted to accelerate convergence. Setting the learning rate to 0.001 can avoid excessive step size causing model oscillation. Setting the number of iterations to 1000 can ensure that the model fully learns the data patterns.

[0080] LSTM models, through deep learning of training data, can more accurately capture the changing patterns of resource demand, with controllable prediction errors, effectively avoiding resource waste or service unavailability caused by prediction bias. Its model structure (multi-hidden-layer design) and training logic (based on multi-feature input and sufficient iterative training) can better fit the sudden characteristics of user requests in edge computing scenarios, ensuring the matching degree between resource supply and actual demand.

[0081] S3032. Normalize the historical request data to obtain normalized data.

[0082] Normalization is used to avoid the influence of different units on the prediction results output by the target prediction model.

[0083] S3033. Use the target prediction model to process the normalized data to obtain the initial resource requirements.

[0084] The target prediction model is used to predict resource demand for the next time period based on historical request data. The initial resource demand is the resource demand output by the target prediction model. Specifically, normalized data is input into the target prediction model, which processes the normalized data to obtain the initial resource demand for the next period.

[0085] S3034. Based on the network quality index, the initial resource requirements are corrected to obtain the target resource requirements.

[0086] Among them, the network quality index is inversely proportional to the target resource requirement. The initial resource requirement can be adjusted by combining the real-time NQI. For example, when the network deteriorates, it may lead to slower data transmission and require more resource buffering. The predicted resource requirement can be increased by 20% as the target resource requirement. For example, if the predicted requirement is 100 CPU cores, it can be adjusted to 120 CPU cores when the NQI drops.

[0087] In the above method, the resources required for the next cycle are predicted by the target prediction model, and the prediction results are dynamically adjusted by real-time NQI to adapt the predicted values ​​to the dynamic state of the network, so as to avoid the imbalance of resource supply and demand caused by network fluctuations and improve the scenario adaptability of the prediction results.

[0088] S304. Based on the target resource requirements and network quality index, classify the active physical hosts to obtain the classification results.

[0089] The classification results are used to initially determine which active physical hosts' virtual machines (VMs) should be migrated out. The classification results can include two categories: active physical hosts from which VMs need to be migrated out, and active physical hosts from which VMs need to be received. The target resource requirements include multiple target user requests. These target user requests can be assigned to the active physical hosts with the highest resource matching degree, ensuring that these active physical hosts are not shut down, thereby allowing VMs on other active physical hosts that do not match target user requests to be migrated out.

[0090] For example, such as Figure 5 As shown, step S304 can be achieved by the following steps:

[0091] S3041. Sort active physical hosts according to their resource utilization, network quality index, and hardware performance.

[0092] The system can normalize resource utilization, network quality index, and hardware performance of active physical hosts. Then, it can weight and sum the normalized data to determine the score for each active physical host, and sort the hosts according to their scores from highest to lowest or lowest to highest. The weights of each item can be flexibly set and adjusted as needed.

[0093] S3042. Sort at least two target user requests according to the resource demand and request priority corresponding to the target resource demand.

[0094] The resource demand and request priority can be normalized, and then the normalized data can be weighted and summed to determine the score corresponding to the user request. The target user requests can then be sorted according to the score from high to low or from low to high.

[0095] Furthermore, the ordering of active physical hosts and target user requests should be done in the same way.

[0096] S3043. The sorted target user requests are sequentially assigned to the sorted active physical hosts to obtain the assignment results.

[0097] The highest-scoring target user request is assigned to the active physical host corresponding to that highest-scoring request. This process is repeated for each target user request, assigning a corresponding active physical host to each request. If all active physical hosts have target user requests assigned to them, but some remain unassigned, the allocation can begin from the active physical host corresponding to the highest score and proceed in a loop to ensure that all target user requests are assigned to their corresponding active physical hosts.

[0098] S3044. Based on the allocation results, active physical hosts that are not assigned to the target user's request are identified as cold physical hosts, and active physical hosts that are assigned to the target user's request are identified as hot physical hosts, thus obtaining the classification results.

[0099] A cold physical host is an active physical host that may not need to process user requests in the next time period, while a hot physical host is an active physical host that may need to process user requests in the next time period. The resource utilization, NQI performance, or hardware performance of a cold physical host are lower than the corresponding indicators of a hot physical host.

[0100] In the above method, target user requests with high resource demand and high priority are preferentially allocated to active physical hosts with good network status and strong processing capabilities. This can avoid response delays or service terminal issues caused by initial allocation to low-performance or poor physical hosts, while avoiding resource idleness caused by allocating target user requests with low resource demand to high-performance physical hosts, thus reducing subsequent virtual machine migration costs.

[0101] S305. Based on the classification results, determine at least two migration candidate schemes, and use a multi-objective optimization model to process the migration candidate schemes, and determine the virtual machine migration strategy from the at least two migration candidate schemes.

[0102] The migration candidate scheme is to migrate virtual machines from cold physical hosts to hot physical hosts based on the classification results. Multiple migration candidate schemes can be included. Furthermore, a multi-objective optimization model can be used to evaluate multiple migration candidate schemes, and the optimal scheme can be selected as the final virtual machine migration strategy.

[0103] For example, the Multi-Objective Two-Stage Variable Neighborhood Searching (MO-TSVNS) algorithm determines the virtual machine migration strategy. In the first stage, a Variable Neighborhood Searching (VNS) strategy is used to identify multiple migration candidate schemes. In the second stage, a multi-objective optimization model is used to determine the optimal scheme from the multiple migration candidate schemes as the final virtual machine migration strategy. Specifically, as... Figure 6 As shown, step S305 can be achieved by the following steps:

[0104] S3051. Based on the classification results, determine at least two migration candidate schemes through a variable neighborhood search strategy.

[0105] The variable neighborhood search strategy includes at least one of the following: a translation strategy, a swap strategy, and a replacement strategy. The translation strategy is used to indicate the migration of virtual machines on a cold physical host to a hot physical host. The swap strategy is used to indicate the exchange of virtual machines on a cold physical host and virtual machines on a hot physical host. The replacement strategy is used to indicate the replacement of virtual machines on a hot physical host with virtual machines on a cold physical host.

[0106] Multiple virtual machine migration schemes between cold and hot physical hosts can be generated using any one or more of the above strategies, resulting in at least two migration candidate schemes. In the translation strategy, virtual machines on any cold or hot physical host can be migrated to any hot physical host. Similarly, in the swap and replacement strategies, the combination of cold and hot physical hosts used for virtual machine swapping or replacement can be randomly determined, meaning the migration candidate schemes include multiple possible combinations.

[0107] S3052. Process the migration candidate schemes through a multi-objective optimization model to obtain the objective function values ​​corresponding to the first and second optimization objectives for the migration candidate schemes.

[0108] The objective function value corresponding to each migration candidate scheme is determined separately. The objective function value is as shown in the formula (2).

[0109] In one possible implementation, the multi-objective optimization model includes a first optimization objective and a second optimization objective, wherein the objective function value corresponding to the first optimization objective is... The objective function value corresponding to the second optimization objective is or .

[0110] In one possible implementation, the multi-objective optimization model includes a first optimization objective, a second optimization objective, and a third optimization objective, where the objective function value corresponding to the first optimization objective is... The objective function value corresponding to the second optimization objective is The objective function value corresponding to the third optimization objective is .

[0111] S3053. Based on the objective function value, select a migration candidate scheme from at least two migration candidate schemes as the virtual machine migration strategy using a non-dominated sorting algorithm and / or a congestion algorithm.

[0112] The non-dominated sorting algorithm selects at least one migration candidate from multiple migration candidates by retaining solutions that are not dominated by other schemes. Then, by calculating the congestion degree, it retains the solutions that are evenly distributed among the above at least one candidate scheme to obtain the optimal scheme, that is, the scheme that best balances energy consumption, response time and reliability, which is used as the final virtual machine migration strategy.

[0113] The above methods comprehensively cover various matching relationships between virtual machines and active physical hosts in a physical machine cluster through migration, exchange, and replacement strategies, providing a sufficient and diverse sample for subsequent screening. The non-dominated sorting algorithm retains solutions for which no other solution is better on the target, avoiding sacrificing overall performance for the sake of a single optimal goal, and ensuring a balance between energy consumption, service quality, and reliability. The congestion algorithm retains evenly distributed solutions, avoiding the concentration of optimal solutions in a certain target range, and ensuring that the final virtual machine migration strategy can adapt to dynamic needs. For example, when network quality deteriorates, a solution with high reliability can be prioritized, and when user requests surge, a solution with low response time can be prioritized.

[0114] S306. According to the virtual machine migration policy, migrate at least one virtual machine from the target active physical host to shut down the target active physical host.

[0115] The target active physical host is one or more of the cold physical hosts. The virtual machines in the target active physical host are migrated to the hot physical host. It is determined whether the resource utilization of the target active physical host is less than a preset threshold. When the utilization of the hot physical host meets the constraint conditions, the target active physical host is shut down.

[0116] For example, such as Figure 7 As shown, step S306 can be achieved by the following steps:

[0117] S3061. Identify the target active physical host among the cold physical hosts.

[0118] The target active physical host is the active physical host that needs to be migrated out as indicated in the migration policy. The target active physical host should be one or more of the cold physical hosts.

[0119] S3062. Back up the data of the virtual machines in the target active physical host to the distributed storage system.

[0120] Distributed storage systems, such as other physical hosts in a physical machine cluster that do not perform virtual machine migration, or remote data backup centers, can avoid data loss during migration if virtual machine data is backed up in advance. For example, if migration fails, the virtual machine can be restarted from backup within 10 seconds.

[0121] S3063. Set up a temporary proxy node between the hot physical host and the target active physical host. The temporary proxy node is used to temporarily handle user requests during the migration.

[0122] Temporary proxy nodes can be physical hosts that do not require virtual machine migration. They are used to temporarily handle user requests during the migration process. Once the migration is complete, the nodes are switched back to the original physical host to ensure the continuity of user services and avoid request failures.

[0123] S3064. According to the virtual machine migration strategy, first migrate the basic data of the virtual machines in the target active physical host to the hot physical host, and then migrate the incremental data of the virtual machines in the target active physical host to the hot physical host.

[0124] Basic data includes at least the initial image data of the virtual machine, and incremental data includes at least temporary files generated by user requests. By transmitting virtual machine data in stages, the total amount of data transmitted in a single transmission is reduced, thereby improving the success rate of virtual machine migration.

[0125] S3065. When the resource utilization rate of the target active physical host is less than the preset second resource utilization rate threshold within a preset time period, and the resource utilization rate of the hot physical host is less than the first resource utilization rate threshold, shut down the target active physical host.

[0126] The second resource utilization threshold is a pre-set threshold used to determine whether to shut down the target active physical host. For example, if the utilization rate of the target active physical host is less than 30% for three consecutive periods, the target active physical host can be shut down.

[0127] Furthermore, the hot physical hosts in the physical machine cluster should meet the resource constraints in the multi-objective optimization model, that is, the utilization rate of the hot physical host receiving virtual machines should be less than the first resource utilization threshold, such as 90%, to ensure that the hot physical host receiving virtual machines can work normally and avoid overloading other physical hosts due to shutting down too many target active physical hosts.

[0128] Among the above methods, building rapid fault recovery capabilities through data backup can improve the reliability of virtual machine migration; setting up temporary agents can achieve seamless service switching and ensure uninterrupted service during migration; migrating basic data and incremental data in batches can reduce network load, avoid transmission timeouts caused by network bandwidth fluctuations, and indirectly reduce the latency of other services; determining whether to shut down target active physical hosts by using thresholds for each active physical host can accurately locate physical hosts that have been idle for a long time without core services, while ensuring that the migrated hot physical hosts will not experience a surge in response time or hardware failure due to excessive load, thus ensuring the stability and reliability of services.

[0129] The server release method provided in this application reduces service interruptions and enhances the system's dynamic network adaptability by real-time sensing of network status and adjusting migration strategies. Through multi-objective optimization, it reduces energy consumption and user request response time, improving migration reliability. The LSTM model predicts resource demands with a prediction error within 10%, demonstrating strong adaptability to sudden requests and reducing resource waste. Targeted migration strategies reduce idle rates on high-configuration nodes and overload rates on low-configuration nodes, improving heterogeneous hardware utilization. Data backup ensures no data loss, improving service continuity and user experience.

[0130] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method.

[0131] Figure 8 This is a schematic diagram of the server release device provided in an embodiment of this application. Figure 8 As shown, embodiments of this application also provide a server release device 80, comprising:

[0132] The acquisition module 801 is used to acquire network status parameters of at least two physical hosts in the physical machine cluster, and determine the network quality index based on the network status parameters. The network quality index is used to indicate the active and inactive physical hosts in the physical machine cluster.

[0133] The first processing module 802 is used to obtain a multi-objective optimization model and generate a virtual machine migration strategy based on the multi-objective optimization model and the network quality index. The optimization objectives of the multi-objective optimization model include at least a first optimization objective and a second optimization objective. The first optimization objective is used to minimize the number of active physical hosts in the physical machine cluster, and the second optimization objective is used to minimize the response time of user requests to active physical hosts in the physical machine cluster, or to maximize the success rate of virtual machine migration between physical hosts in the physical machine cluster.

[0134] The second processing module 803 is used to migrate at least one virtual machine from the target active physical host according to the virtual machine migration policy, so as to shut down the target active physical host.

[0135] In one possible implementation, the multi-objective optimization model includes at least a first optimization objective, a second optimization objective, and a third optimization objective. The model parameters corresponding to the optimization objectives satisfy at least one of the following constraints: resource constraints, migration time constraints, and compatibility constraints.

[0136] The first optimization objective is to minimize energy consumption based on the number of active physical hosts and the total number of physical hosts.

[0137] The second optimization objective is to minimize the response time of user requests based on the processing capacity of the active physical host and the distance between the user and the active physical host.

[0138] The third optimization objective is to maximize the success rate of virtual machine migration between physical hosts within a physical machine cluster based on historical migration data. Historical migration data includes the number of migration failures, data loss, and total data volume during virtual machine migration in the previous period.

[0139] Resource constraints are used to limit the resource utilization of active physical hosts after virtual machine migration to no more than a first resource utilization threshold.

[0140] Migration time constraints are used to limit the migration time when migrating virtual machines to no more than the response time threshold.

[0141] Compatibility constraints are used to ensure that virtual machines match the operating system and hardware architecture of the active physical host.

[0142] In one possible implementation, module 801 is specifically used for:

[0143] The network status of the physical host is collected in real time according to the preset collection frequency to obtain network status parameters, which include at least one of the following: network bandwidth, end-to-end latency, physical host movement speed, and signal strength.

[0144] The network quality index is obtained by weighted summation of network status parameters. The network quality index of active physical hosts is better than that of inactive physical hosts.

[0145] In one possible implementation, the first processing module 802 is specifically used for:

[0146] Obtain historical request data, which includes at least one of the following: request type, resource requirement, and request arrival time.

[0147] Based on historical request data and network quality index, the resource demand for the next time period is predicted to obtain the target resource demand. The target resource demand includes at least two target user requests, as well as the resource demand and request priority corresponding to the target user requests.

[0148] The virtual machine migration strategy is determined based on the multi-objective optimization model, target resource requirements, and network quality index.

[0149] In one possible implementation, the first processing module 802 is further configured to:

[0150] Obtain the target prediction model, which is a long short-term memory network model;

[0151] The historical request data is normalized to obtain normalized data;

[0152] The target prediction model is used to process the normalized data to obtain the initial resource requirements, and the target detection model is used to predict the resource requirements in the next time period based on historical request data.

[0153] Based on the network quality index, the predicted resource demand is corrected to obtain the target resource demand, where the network quality index is inversely proportional to the target resource demand.

[0154] In one possible implementation, the first processing module 802 is further configured to:

[0155] Based on the target resource requirements and network quality index, active physical hosts are classified to obtain the classification results;

[0156] Based on the classification results, at least two migration candidate schemes are identified, and a multi-objective optimization model is used to process the migration candidate schemes to determine the virtual machine migration strategy from the at least two migration candidate schemes.

[0157] In one possible implementation, the first processing module 802 is further configured to:

[0158] Active physical hosts are ranked based on their resource utilization, network quality index, and hardware performance.

[0159] Based on the resource demand and request priority corresponding to the target resource demand, sort at least two target user requests;

[0160] The sorted target user requests are then sequentially assigned to the sorted active physical hosts to obtain the allocation results.

[0161] Based on the allocation results, active physical hosts that were not assigned to the target user's request are identified as cold physical hosts, and active physical hosts that were assigned to the target user's request are identified as hot physical hosts, thus obtaining the classification results.

[0162] In one possible implementation, the first processing module 802 is further configured to:

[0163] Based on the classification results, at least two migration candidate schemes are determined through a variable neighborhood search strategy. The variable neighborhood search strategy includes at least one of the following: a translation strategy, a swap strategy, and a replacement strategy. The translation strategy is used to indicate the migration of virtual machines on cold physical hosts to hot physical hosts. The swap strategy is used to indicate the swapping of virtual machines on cold physical hosts with virtual machines on hot physical hosts. The replacement strategy is used to indicate the replacement of virtual machines on hot physical hosts with virtual machines on cold physical hosts.

[0164] By processing migration candidate schemes using a multi-objective optimization model, the objective function values ​​corresponding to the first and second optimization objectives for the migration candidate schemes are obtained.

[0165] Based on the objective function value, a migration candidate is selected from at least two migration candidates as the virtual machine migration strategy using a non-dominated sorting algorithm and / or a crowding algorithm.

[0166] In one possible implementation, the second processing module 803 is specifically used for:

[0167] Identify the target active physical host within the cold physical host;

[0168] Back up the data of virtual machines on the target active physical host to a distributed storage system;

[0169] A temporary proxy node is set up between the hot physical host and the target active physical host. The temporary proxy node is used to temporarily handle user requests during the migration.

[0170] According to the virtual machine migration strategy, the basic data of the virtual machines in the target active physical host is first migrated to the hot physical host, and then the incremental data of the virtual machines in the target active physical host is migrated to the hot physical host. The basic data includes at least the initial image data of the virtual machines, and the incremental data includes at least the temporary files generated by user requests.

[0171] When the resource utilization rate of the target active physical host is less than the preset second resource utilization rate threshold within a preset time period, and the resource utilization rate of the hot physical host is less than the first resource utilization rate threshold, the target active physical host is shut down.

[0172] For a description of the features in the embodiment corresponding to the server release device 80, please refer to the relevant description in the embodiment corresponding to the server release method, which will not be repeated here.

[0173] Figure 9 A schematic diagram of the structure of the electronic device provided in this application. Figure 9 As shown, the electronic device 90 provided in this embodiment includes at least one processor 901 and a memory 902. Optionally, the electronic device 90 further includes a communication component 903. The processor 901, memory 902, and communication component 903 are connected via a bus.

[0174] In a specific implementation, at least one processor 901 executes computer execution instructions stored in memory 902, causing at least one processor 901 to execute the above-described server release method embodiment.

[0175] The specific implementation process of processor 901 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0176] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the application can be directly manifested as being executed by a hardware processor, or executed by a combination of hardware and software modules within the processor.

[0177] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0178] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0179] Embodiments of this application also provide a computer-readable storage medium storing a computer program, wherein the computer program is configured to execute the steps in any of the above-described server release method embodiments when it runs.

[0180] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard disk, magnetic disk, or optical disk.

[0181] Embodiments of this application also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the above-described server release method embodiments.

[0182] Embodiments of this application also provide another computer program product, including a non-volatile computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps in any of the above-described server release method embodiments.

[0183] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0184] The server release method provided in this application has been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only intended to help understand the method and core ideas of this application. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of the claims of this application.

Claims

1. A server release method, characterized in that, include: Obtain network status parameters of at least two physical hosts in the physical machine cluster, and determine a network quality index based on the network status parameters. The network quality index is used to indicate active and inactive physical hosts in the physical machine cluster. A multi-objective optimization model is obtained, and a virtual machine migration strategy is generated based on the multi-objective optimization model and the network quality index. The optimization objectives of the multi-objective optimization model include at least a first optimization objective and a second optimization objective. The first optimization objective is used to minimize the number of active physical hosts in the physical machine cluster, and the second optimization objective is used to minimize the response time of user requests to active physical hosts in the physical machine cluster, or to maximize the success rate of virtual machine migration between physical hosts in the physical machine cluster. According to the virtual machine migration policy, at least one virtual machine in the target active physical host is migrated out in order to shut down the target active physical host; The multi-objective optimization model has at least three optimization objectives: a first optimization objective, a second optimization objective, and a third optimization objective. The model parameters corresponding to these optimization objectives satisfy at least one of the following constraints: resource constraints, migration time constraints, and compatibility constraints. The first optimization objective is to minimize energy consumption based on the number of active physical hosts and the total number of physical hosts; The second optimization objective is to minimize the response time of user requests based on the processing capacity of the active physical host and the distance between the user and the active physical host; The third optimization objective is to maximize the success rate of virtual machine migration between physical hosts in the physical machine cluster based on historical migration data. The historical migration data includes the number of migration failures, data loss, and total data volume during virtual machine migration in the previous period. The resource constraint is used to limit the resource utilization of the active physical host after the virtual machine migration to not exceed the first resource utilization threshold. The migration time constraint is used to limit the migration time during virtual machine migration to no more than the response time threshold. The compatibility constraints are used to ensure that the virtual machine matches the operating system and hardware architecture of the active physical host.

2. The server release method according to claim 1, characterized in that, The step of obtaining network status parameters of at least two physical hosts in the physical machine cluster and determining the network quality index based on the network status parameters includes: The network status of the physical host is collected in real time according to a preset collection frequency to obtain the network status parameters, which include at least one of the following: network bandwidth, end-to-end latency, physical host movement speed, and signal strength. The network quality index is obtained by weighted summation of the network status parameters, and the network quality index of the active physical host is better than that of the inactive physical host.

3. The server release method according to claim 1, characterized in that, The process of generating a virtual machine migration strategy based on the multi-objective optimization model and the network quality index includes: Obtain historical request data, which includes at least one of the following: request type, resource requirement, and request arrival time; Based on the historical request data and the network quality index, the resource demand in the next time period is predicted to obtain the target resource demand. The target resource demand includes at least two target user requests, as well as the resource demand and request priority corresponding to the target user requests. The virtual machine migration strategy is determined based on the multi-objective optimization model, the target resource requirements, and the network quality index.

4. The server release method according to claim 3, characterized in that, The step of predicting resource demand for the next time period based on the historical request data and the network quality index to obtain the target resource demand includes: Obtain the target prediction model, which is a long short-term memory network model; The historical request data is normalized to obtain normalized data; The normalized data is processed using the target prediction model to obtain the initial resource requirements, and the target detection model is used to predict the resource requirements for the next time period based on historical request data. Based on the network quality index, the initial resource requirement is modified to obtain the target resource requirement, wherein the network quality index is inversely proportional to the target resource requirement.

5. The server release method according to claim 3, characterized in that, The step of determining the virtual machine migration strategy based on the multi-objective optimization model, the target resource requirements, and the network quality index includes: Based on the target resource requirements and the network quality index, the active physical hosts are classified to obtain the classification results; Based on the classification results, at least two migration candidate schemes are determined, and the multi-objective optimization model is used to process the migration candidate schemes to determine the virtual machine migration strategy from the at least two migration candidate schemes.

6. The server release method according to claim 5, characterized in that, The process of classifying active physical hosts based on the target resource requirements and the network quality index to obtain classification results includes: The active physical hosts are ranked according to their resource utilization, network quality index, and hardware performance. Based on the resource demand and request priority corresponding to the target resource demand, sort at least two target user requests; The sorted target user requests are sequentially assigned to the sorted active physical hosts to obtain the assignment results; Based on the allocation results, active physical hosts that are not assigned to the target user's request are identified as cold physical hosts, and active physical hosts assigned to the target user's request are identified as hot physical hosts, thus obtaining the classification results.

7. The server release method according to claim 6, characterized in that, The step of determining at least two migration candidate schemes based on the classification results, analyzing and processing the migration candidate schemes using the multi-objective optimization model, and determining the virtual machine migration strategy from the at least two migration candidate schemes includes: Based on the classification results, at least two migration candidate schemes are determined through a variable neighborhood search strategy. The variable neighborhood search strategy includes at least one of the following: a translation strategy, a swap strategy, and a replacement strategy. The translation strategy is used to indicate the migration of virtual machines on cold physical hosts to hot physical hosts. The swap strategy is used to indicate the swapping of virtual machines on cold physical hosts with virtual machines on hot physical hosts. The replacement strategy is used to indicate the replacement of virtual machines on hot physical hosts with virtual machines on cold physical hosts. The migration candidate schemes are processed by the multi-objective optimization model to obtain the objective function values ​​corresponding to the first and second optimization objectives for the migration candidate schemes. Based on the objective function value, a migration candidate scheme is selected from at least two migration candidate schemes as the virtual machine migration strategy using a non-dominated sorting algorithm and / or a crowding algorithm.

8. The server release method according to any one of claims 1 to 7, characterized in that, The step of migrating at least one virtual machine from a target active physical host according to the virtual machine migration policy, in order to shut down the target active physical host, includes: Identify the target active physical host from the cold physical hosts; Back up the data of the virtual machines within the target active physical host to a distributed storage system; A temporary proxy node is set up between the thermal physical host and the target active physical host. The temporary proxy node is used to temporarily handle user requests during the migration. According to the virtual machine migration strategy, the basic data of the virtual machines in the target active physical host is first migrated to the hot physical host, and then the incremental data of the virtual machines in the target active physical host is migrated to the hot physical host. The basic data includes at least the initial image data of the virtual machines, and the incremental data includes at least the temporary files generated by user requests. When the resource utilization rate of the target active physical host is less than the preset second resource utilization rate threshold within a preset time period, and the resource utilization rate of the hot physical host is less than the first resource utilization rate threshold, the target active physical host is shut down.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the server release method as described in any one of claims 1 to 8 when executing the computer program.

Images