Thermal Migration Result Prediction Method, Thermal Migration Method, Electronic Device, and Storage Medium

Abstract

The present disclosure provides a method for predicting hot migration results, a hot migration method, an electronic device, and a storage medium, relating to cloud server technology, including: obtaining historical performance data of a virtual machine within a preset time period every day in the past N days and static attribute information of the virtual machine; determining feature information of the virtual machine within the preset time period according to the historical performance data and the static attribute information; inputting the feature information into a preset prediction model to obtain the success rate when hot migrating the virtual machine in a future preset time period, and the success rate is used to characterize the predicted hot migration result of the virtual machine. In this implementation manner, it is possible to predict the success rate of hot migrating the virtual machine in a future preset time period by using the historical data of the virtual machine within the preset time period. This method not only considers the static characteristics of the virtual machine but also the dynamic characteristics of the virtual machine within different preset time periods. Therefore, the solution provided by the present disclosure can leave enough time for hot migration planning of the virtual machine.

Description

Technical Field

[0001] The present disclosure relates to cloud server technology, and in particular to a method for predicting hot migration results, a hot migration method, an electronic device, and a storage medium. Background Art

[0002] Currently, many Internet services are deployed in cloud servers. Virtual machines are started in the cloud servers to process services.

[0003] Generally, in order to make full use of the computing resources in cloud servers, there is a virtual machine hot migration technology. Virtual machine hot migration refers to the process of moving a running virtual machine between different physical machines without disconnecting the connection with the client or closing the application program.

[0004] Whether the virtual machine hot migration is successful or not will affect the services being processed by the virtual machine. Therefore, it is necessary to improve the success rate of virtual machine hot migration, so as to improve the stability of the services processed in the cloud server. Summary of the Invention

[0005] The present disclosure provides a method for predicting hot migration results, a hot migration method, an electronic device, and a storage medium, which can predict the hot migration results of a virtual machine, and further plan the hot migration timing of the virtual machine in advance.

[0006] The first aspect of the present disclosure is to provide a method for predicting the hot migration results of a virtual machine, including:

[0007] Obtaining historical performance data of the virtual machine within a preset time period every day in the past N days and static attribute information of the virtual machine; N is an integer greater than or equal to 1;

[0008] Determining feature information of the virtual machine within the preset time period according to the historical performance data and the static attribute information;

[0009] Inputting the feature information into a preset prediction model to obtain the success rate when hot migrating the virtual machine within a preset time period in the future, where the success rate is used to represent the predicted migration result of the virtual machine.

[0010] The second aspect of the present disclosure is to provide a hot migration method for a virtual machine, including:

[0011] Determining the success rate when hot migrating a target virtual machine within a preset time period in the future according to the method described in the first aspect, and determining the migration time according to the success rate;

[0012] Hot migrating the target virtual machine at the migration time.

[0013] The third aspect of the present disclosure is to provide a method for training a model for predicting the hot migration results of a virtual machine, including:

[0014] Obtain the training performance data of the virtual machine within a preset time period every day in N days, the static attribute information of the virtual machine, and the migration result information of the virtual machine within a preset time period after N days;

[0015] Determine the training sample information of the virtual machine within the preset time period according to the training performance data and the static attribute information;

[0016] Input the training sample information into a preset model to obtain the success rate when hot migrating the virtual machine within a preset time period after N days;

[0017] Adjust the parameters in the preset model according to the success rate and the migration result information of the virtual machine within a preset time period after N days, where the preset model after meeting the stop training condition is a model for predicting the hot migration result of the virtual machine. The fourth aspect of the present disclosure is to provide a device for predicting the hot migration result of a virtual machine, including:

[0018] An acquisition unit, configured to acquire the historical performance data of the virtual machine within a preset time period every day in the past N days and the static attribute information of the virtual machine; N is an integer greater than or equal to 1;

[0019] A feature determination unit, configured to determine the feature information of the virtual machine within the preset time period according to the historical performance data and the static attribute information;

[0020] A prediction unit, configured to input the feature information into a preset prediction model to obtain the success rate when hot migrating the virtual machine within a preset time period in the future, where the success rate is used to characterize the predicted migration result of the virtual machine.

[0021] The fifth aspect of the present disclosure is to provide a hot migration device for a virtual machine, including:

[0022] A migration time determination unit, configured to determine the success rate when hot migrating a target virtual machine within a preset time period in the future according to the device described in the fourth aspect, and determine the migration time according to the success rate;

[0023] A hot migration unit, configured to hot migrate the target virtual machine at the migration time.

[0024] The sixth aspect of the present disclosure is to provide a training device for a model for predicting the hot migration result of a virtual machine, including:

[0025] An acquisition unit, configured to acquire the training performance data of the virtual machine within a preset time period every day in N days, the static attribute information of the virtual machine, and the migration result information of the virtual machine within a preset time period after N days;

[0026] A sample determination unit, configured to determine training sample information of the virtual machine in the preset time period according to the training performance data and the static attribute information;

[0027] A prediction unit, configured to input the training sample information into a preset model to obtain a success rate when the virtual machine is hot migrated in a preset time period N days later;

[0028] An adjustment unit, configured to adjust parameters in the preset model according to the success rate and migration result information of the virtual machine in the preset time period N days later, where the preset model after meeting the stop training condition is a model for predicting the hot migration result of the virtual machine.

[0029] The seventh aspect of the present disclosure is to provide an electronic device, including:

[0030] A memory;

[0031] A processor; and

[0032] A computer program;

[0033] Wherein, the computer program is stored in the memory and is configured to be executed by the processor to implement the method described in the first aspect or the second aspect.

[0034] The eighth aspect of the present disclosure is to provide a computer-readable storage medium, on which a computer program is stored, and the computer program is executed by a processor to implement the method described in the first aspect, the second aspect or the third aspect.

[0035] The technical effects of the hot migration result prediction method, the hot migration method, the electronic device and the storage medium provided by the present disclosure are:

[0036] The hot migration result prediction method, the hot migration method, the electronic device and the storage medium provided in this embodiment include: obtaining historical performance data of the virtual machine in a preset time period every day in the past N days and static attribute information of the virtual machine; N is an integer greater than or equal to 1; determining feature information of the virtual machine in the preset time period according to the historical performance data and the static attribute information; inputting the feature information into a preset prediction model to obtain a success rate when the virtual machine is hot migrated in a future preset time period, where the success rate is used to characterize the predicted hot migration result of the virtual machine. In this implementation manner, the success rate of hot migrating the virtual machine in a future preset time period can be predicted by using the historical data of the virtual machine in the preset time period. This method not only considers the static characteristics of the virtual machine but also considers the dynamic characteristics of the virtual machine in different preset time periods. Therefore, the solution provided by the present disclosure can leave enough time for the hot migration planning of the virtual machine. Description of the Drawings

[0037] Figure 1 A scenario diagram shown in an exemplary embodiment of the present disclosure;

[0038] Figure 2 A flowchart of a method for predicting the result of virtual machine live migration shown in an exemplary embodiment of the present application;

[0039] Figure 3 A flowchart of a method for predicting the result of virtual machine live migration shown in another exemplary embodiment of the present application;

[0040] Figure 4 A schematic diagram of the live migration prediction result of a virtual machine shown in an exemplary embodiment of the present disclosure;

[0041] Figure 5 A schematic diagram of a solution for determining the live migration success rate shown in an exemplary embodiment of the present disclosure;

[0042] Figure 6 A schematic flowchart of a method for training a model for predicting the result of virtual machine live migration shown in an exemplary embodiment of the present disclosure;

[0043] Figure 7 A structural diagram of a virtual machine live migration result prediction device shown in an exemplary embodiment of the present application;

[0044] Figure 8 A structural diagram of a virtual machine live migration result prediction device shown in another exemplary embodiment of the present application;

[0045] Figure 9 A structural diagram of a training device for a model for predicting the result of virtual machine live migration shown in an exemplary embodiment of the present application;

[0046] Figure 10 A structural diagram of an electronic device shown in an exemplary embodiment of the present application. Detailed implementation manners

[0047] Figure 1 A scenario diagram shown in an exemplary embodiment of the present disclosure.

[0048] As Figure 1 shown, multiple cloud servers are set in the cloud service platform. Virtual machines can be started in the cloud servers, and the virtual machines are used to process services, so that multiple services can be processed simultaneously in one cloud server, and the services are isolated from each other.

[0049] When the cloud service platform is running, there is a need for live migration of virtual machines.

[0050] For example, when the cloud server 11 needs to be repaired to improve its stability, the virtual machines in the cloud server 11 need to be hot migrated to other cloud servers, such as being migrated to the cloud server 12, so as not to interrupt the service, enabling the cloud server for processing services to be replaced without the user's awareness.

[0051] For another example, in order to reduce costs, it is usually necessary to reduce the generation of virtual machine CPU fragmentation. Therefore, it is necessary to reasonably allocate virtual machine CPUs through hot migration technology. For example, it is necessary to create a virtual machine with a 4-core CPU in the cloud server 13. Multiple virtual machines have been started in the cloud server 13, leaving only 2 cores of available CPU in this cloud server. At this time, the virtual machines in the cloud server 13 need to be hot migrated to other cloud servers, such as being migrated to the cloud server 14.

[0052] To improve the success rate of virtual machine migration, the migration recommendations for virtual machines are usually determined based on the real-time data of the virtual machines, such as recommending migration or not recommending migration. However, this solution requires the use of the real-time data of the virtual machines, so the migration timing of the virtual machines cannot be planned in advance.

[0053] To solve the above technical problems, in the solution provided by the present disclosure, the historical performance data of the virtual machine within a certain preset time period and the static attribute information of the virtual machine are used to predict the migration success rate of the virtual machine in the future within the preset time period. Furthermore, migration recommendations can be provided based on the migration success rate of the virtual machine. This implementation solution can not only improve the migration success rate of the virtual machine but also plan the migration timing of the virtual machine in advance.

[0054] Figure 2 It is a flowchart of a virtual machine hot migration result prediction method shown in an exemplary embodiment of the present application.

[0055] As Figure 2 shown, the virtual machine hot migration result prediction method provided in this embodiment includes:

[0056] Step 201, obtain the historical performance data of the virtual machine within the preset time period every day in the past N days and the static attribute information of the virtual machine; N is an integer greater than or equal to 1.

[0057] In the solution provided by the present disclosure, it can be executed by an electronic device with computing capabilities or by a system including multiple electronic devices. The electronic device can be, for example, a server. A virtual machine can be set in this server, and the server or system can predict the hot migration success rate of the virtual machine in the future within the preset time period. If this method is executed by the server, a virtual machine may not be set in this server, and the server can predict the hot migration success rate of the virtual machines set in other servers in the future within the preset time period.

[0058] Among them, when it is necessary to plan the hot migration timing for a virtual machine, the electronic device can obtain the past historical performance data of the virtual machine, and can also obtain the static attribute information of the virtual machine. The electronic device can use the relevant data of the virtual machine to predict the hot migration success rate of the virtual machine within a preset time period in the future.

[0059] Specifically, for example, if it is necessary to predict whether the hot migration of a virtual machine can succeed within a certain preset time period in the future, the historical performance data of the virtual machine within the preset time period every day in the past N days can be obtained. For example, if the current time is the T-th day and it is necessary to predict the success rate of hot migrating the virtual machine from 9:00 to 10:00 on the (T + 1)-th day, then the historical performance data of the virtual machine from 9:00 to 10:00 every day within the period from the (T - N - 1)-th day to the (T - 1)-th day can be obtained on the T-th day.

[0060] The historical performance data refers to the performance data when the virtual machine is running, which is dynamic data. Since the business peak hours are different, the historical performance data of the virtual machine within different preset time periods is also different. Based on the running conditions of the virtual machine in different time periods, the success rate of hot migrating the virtual machine in different time periods can be accurately predicted.

[0061] The static attribute information is the static data of the virtual machine and does not change with the change of the business. The static attributes of the virtual machine will also affect the hot migration success rate of the virtual machine. Therefore, in the solution provided by the present disclosure, the static attributes of the virtual machine and the dynamic attributes of the virtual machine within a certain preset time period can be combined to predict the hot migration success rate of the virtual machine within the preset time period in the future.

[0062] Optionally, N can be an integer greater than or equal to 1, such as 7. It can be specifically set according to requirements.

[0063] Among them, the historical performance data may include the CPU usage, memory usage, and data input and output conditions when the virtual machine is running, and the static attribute information may include data such as the CPU data, memory data, and model of the virtual machine.

[0064] Step 202: Determine the feature information of the virtual machine within the preset time period according to the historical performance data and the static attribute information.

[0065] Specifically, the electronic device can process the historical performance data and the static attribute information of the virtual machine within the preset time period to obtain the feature information of the virtual machine within the preset time period.

[0066] Further, the historical performance data can be data corresponding to multiple moments. For example, it can be data at each moment within a preset time period for each day in N days of a virtual machine. Then, based on multiple pieces of the same type of performance data in the historical performance data, performance characteristics can be obtained. By combining multiple performance characteristics and static attribute information, the characteristic information of the virtual machine within the preset time period can be obtained.

[0067] In practical applications, for example, in the historical performance data, there can be multiple pieces of data related to CPU usage. For example, within a preset time period for each day in the past N days, there are data related to CPU usage corresponding to M moments. Then, there are M * N pieces of data related to CPU usage in the historical performance data. These M * N pieces of data related to CPU usage can be processed to obtain the performance characteristics related to CPU usage of the virtual machine within the preset time period.

[0068] Among them, for each type of historical performance data, it can be processed to obtain the corresponding performance characteristics, which are used to characterize the dynamic characteristics when the virtual machine is running.

[0069] In an optional implementation manner, the respective performance characteristics and static attribute information of the virtual machine can be spliced to obtain the characteristic information of the virtual machine within the preset time period. For each preset time period, the electronic device can determine the characteristic information of the virtual machine. For example, if the duration of each preset time period is 1 hour, the electronic device can determine the characteristic information of the virtual machine for each hour within the next day.

[0070] Step 203: Input the characteristic information into a preset prediction model to obtain the success rate of hot migrating the virtual machine in a future preset time period, where the success rate is used to characterize the predicted hot migration result of the virtual machine.

[0071] Among them, the electronic device inputs the characteristic information of the virtual machine within the preset time period into the preset prediction model, and this prediction model can output the success rate of hot migrating the virtual machine in this preset time period in the future.

[0072] Specifically, this prediction model can be pre-trained. For example, it can be trained using the historical performance data of the virtual machine within a preset time period for each day in N days, the static attribute information of this virtual machine, and the result of hot migrating the virtual machine in this preset time period in the future.

[0073] Further, an offline dataset can be obtained in advance. The offline dataset includes the historical performance data of the virtual machine in multiple preset time periods, and can also include the static attribute information of the virtual machine, and can also include the results of live migration of the virtual machine in each preset time period. Training data and its corresponding labels can be determined based on these data. The training data can be the historical performance data of the virtual machine in the preset time period every day in N days and the static attribute information of the virtual machine, and the label of the training data can be the result of live migration of the virtual machine in the future preset time period.

[0074] In actual application, the trained prediction model can be deployed in an electronic device, so that the electronic device can use the prediction model to predict the predicted live migration result of the virtual machine in the preset time period.

[0075] Among them, the electronic device can also determine whether the live migration of the virtual machine can be successful in the future preset time period according to the success rate output by the prediction model. For example, a threshold can be set. If the success rate is greater than or equal to the set threshold, it can be determined that the live migration of the virtual machine can be successful in the future preset time period. If the success rate is less than the set threshold, it can be determined that the live migration of the virtual machine cannot be successful in the future preset time period.

[0076] The method for predicting the live migration result of a virtual machine provided by the present disclosure includes: obtaining the historical performance data of the virtual machine in the preset time period every day in the past N days and the static attribute information of the virtual machine; N is an integer greater than or equal to 1; determining the characteristic information of the virtual machine in the preset time period according to the historical performance data and the static attribute information; inputting the characteristic information into a preset prediction model to obtain the success rate when live migrating the virtual machine in the future preset time period, where the success rate is used to characterize the predicted live migration result of the virtual machine. In this implementation manner, the success rate of live migrating the virtual machine in the future preset time period can be predicted by using the historical data of the virtual machine in the preset time period. This method takes into account both the static characteristics of the virtual machine and the dynamic characteristics of the virtual machine in different preset time periods. Therefore, the solution provided by the present disclosure can leave enough time for the live migration planning of the virtual machine.

[0077] Figure 3 This is a flowchart of the method for predicting the live migration result of a virtual machine shown in another exemplary embodiment of the present application.

[0078] As Figure 3 shown, the method for predicting the live migration result of a virtual machine provided in this embodiment includes:

[0079] Step 301, obtain the historical performance data of the virtual machine in the preset time period every day in the past N days; N is an integer greater than or equal to 1.

[0080] The implementation of step 301 is similar to that of step 201 and will not be elaborated here.

[0081] Step 302, obtain the first attribute information of the virtual machine itself and the second attribute information of the physical machine where the virtual machine is located.

[0082] Among them, the static attribute information of the virtual machine also has an impact on the successful hot migration of the virtual machine. Therefore, in the method provided by the present disclosure, the static attribute information of the virtual machine can also be obtained, so as to predict the success rate of the hot migration of the virtual machine by combining the static attribute information and the dynamic information of the virtual machine.

[0083] Specifically, the static attribute information of the virtual machine includes the first attribute information of itself and also includes the second attribute information of the physical machine where the virtual machine is located.

[0084] Furthermore, the first attribute information refers to the attributes of the virtual machine itself, and specifically may include at least one of the following: the number of central processing units of the virtual machine, the memory size of the virtual machine, and the operating system information of the virtual machine.

[0085] The second attribute information is the attribute of the physical machine on which the virtual machine is installed, and specifically may include at least one of the following: the number of cores of the central processing unit of the physical machine, the model information of the physical machine, and the network card information of the physical machine.

[0086] Step 303, determine the historical performance characteristics of the virtual machine within a preset time period according to the data corresponding to each moment within the preset time period in the historical performance data.

[0087] Among them, the historical performance data includes the data corresponding to each moment within the preset time period. For example, if there are M moments within the preset time period, then there are data corresponding to the M moments every day within each preset time period. The data at each moment in each preset time period within N days can be processed to determine the historical performance characteristics of the virtual machine within the preset time period.

[0088] For example, if the preset time period is from 8:00 to 9:00 every day, then the historical performance data of the virtual machine from 8:00 to 9:00 every day within N days can be obtained. The historical performance data specifically includes the data corresponding to multiple moments from 8:00 to 9:00, such as the data per second. The data at each moment from 8:00 to 9:00 every day within N days can be processed to obtain the historical performance characteristics of the virtual machine from 8:00 to 9:00.

[0089] Specifically, the historical performance characteristics are used to characterize the characteristics of the virtual machine when it is running within the preset time period.

[0090] Further, the historical performance data has data corresponding to each moment in the time dimension, and each moment's data includes various data related to the performance of the virtual machine. For example, each moment's data can include at least one of the following: the central processing unit (CPU) usage rate of the virtual machine, the memory usage rate of the virtual machine, the network traffic data of the virtual machine, the data read / write times information of the virtual machine, and the memory bandwidth data of the virtual machine.

[0091] In actual application, at different moments, the services processed by the virtual machine are not the same. Therefore, the CPU operation situation, memory operation situation, etc. of the virtual machine are all different.

[0092] Therefore, the operation situation of the virtual machine at different moments can be characterized by the relevant data such as the CPU and memory of the virtual machine, which can specifically include the central processing unit (CPU) usage rate of the virtual machine, the memory usage rate of the virtual machine, the network traffic data of the virtual machine, the data read / write times information of the virtual machine, and the memory bandwidth data of the virtual machine.

[0093] Further, the CPU usage rate of the virtual machine refers to the CPU usage rate at each moment when the virtual machine is running.

[0094] In actual application, the memory usage rate of the virtual machine refers to the memory usage rate at each moment when the virtual machine is running.

[0095] Among them, the network traffic data of the virtual machine refers to the network traffic situation of the virtual machine at each moment.

[0096] Specifically, the data read / write times information of the virtual machine refers to the data input / output data of the virtual machine at each moment when it is running. For example, the amount of data input to the virtual machine at each moment and the amount of data output from the virtual machine at each moment.

[0097] Further, the memory bandwidth data of the virtual machine refers to the size of the memory bandwidth when the virtual machine is running. Memory bandwidth refers to the rate at which the CPU of the virtual machine can read data from the memory or store data in the memory.

[0098] In actual application, if the historical performance data includes the data of the CPU usage rate corresponding to each moment within a preset time period, the historical performance characteristics determined by the electronic device can include:

[0099] Determine the CPU performance characteristics according to the CPU usage rate of the virtual machine corresponding to each moment within the preset time period; the CPU performance characteristics include at least one of the following: the average value of the CPU usage rate, the percentile value of the CPU usage rate, and the variance of the CPU usage rate.

[0100] Among them, if the historical performance data includes the data of the memory usage rate of the virtual machine corresponding to each moment within a preset time period, the historical performance characteristics determined by the electronic device can include:

[0101] Determine the first memory performance characteristic according to the memory utilization rate of the virtual machine corresponding to each moment within a preset time period; the first memory performance characteristic includes at least one of the following: average memory utilization rate, memory utilization rate quantile value, memory utilization rate variance.

[0102] Specifically, if the historical performance data includes the data of the network traffic data of the virtual machine corresponding to each moment within the preset time period, the historical performance characteristics determined by the electronic device may include:

[0103] Determine the network performance characteristic according to the network traffic data of the virtual machine corresponding to each moment within the preset time period; the network performance characteristic includes at least one of the following: average network traffic, network traffic quantile value, network traffic variance.

[0104] Furthermore, if the historical performance data includes the data of the number of data read / write times of the virtual machine corresponding to each moment within the preset time period, the historical performance characteristics determined by the electronic device may include:

[0105] Determine the data read / write performance characteristic according to the number of data read / write times information of the virtual machine corresponding to each moment within the preset time period; the data read / write performance characteristic includes at least one of the following: average number of data read / write times, data read / write times quantile value, data read / write times variance.

[0106] In actual application, if the historical performance data includes the data of the memory bandwidth data of the virtual machine corresponding to each moment within the preset time period, the historical performance characteristics determined by the electronic device may include:

[0107] Determine the second memory performance characteristic according to the memory bandwidth data of the virtual machine corresponding to each moment within the preset time period; the second memory performance characteristic includes at least one of the following: average memory bandwidth, memory bandwidth quantile value, memory bandwidth variance.

[0108] For the CPU utilization rate of the data virtual machine corresponding to each moment within the preset time period in the historical performance data, the memory utilization rate of the virtual machine, the network traffic data of the virtual machine, the number of data read / write times information of the virtual machine, the memory bandwidth data of the virtual machine, etc., corresponding mean values, quantile values, variance values, etc. can be determined, and these are used as the historical performance characteristics of the virtual machine within the preset time period.

[0109] Among them, the quantile value may specifically include the 5th quantile value, 25th quantile value, 50th quantile value, 75th quantile value, 95th quantile value, etc. One or more quantile values can be determined for each type of performance data, and specifically there is no limit to this. For example, one or more of the 5th quantile value, 25th quantile value, 50th quantile value, 75th quantile value, 95th quantile value can be determined for the CPU utilization rate of the virtual machine.

[0110] Step 304: Concatenate the historical performance characteristics of the virtual machine with the static attribute information of the virtual machine to obtain the characteristic information of the virtual machine in a preset time period.

[0111] Further, after the electronic device determines the historical performance characteristics of the virtual machine in the preset time period, it can concatenate the historical performance characteristics of the virtual machine with the static attribute information of the virtual machine, and then obtain the characteristic information of the virtual machine in the preset time period. For example, the historical performance characteristics, the first attribute information, and the second attribute information can be concatenated to obtain the characteristic information of the virtual machine in the preset time period.

[0112] Step 305: Input the characteristic information into a preset prediction model to obtain the success rate of hot migrating the virtual machine in a future preset time period, where the success rate is used to characterize the thermal migration result of the virtual machine.

[0113] The implementation manner of Step 305 is similar to that of Step 203 and will not be elaborated here.

[0114] Step 306: Compare the success rate with a preset threshold, and determine whether the hot migration of the virtual machine can succeed in a future preset time period according to the comparison result.

[0115] In practical applications, the electronic device can compare the success rate output by the prediction model with the preset threshold. If the success rate is greater than or equal to the preset threshold, it can be determined that the hot migration of the virtual machine can succeed in a future preset time period; if the success rate is less than the preset threshold, it can be determined that the hot migration of the virtual machine cannot succeed in a future preset time period.

[0116] The preset threshold can be set according to requirements, and the present disclosure does not limit it.

[0117] Step 307: Determine whether the hot migration of the virtual machine can succeed in multiple preset time periods in the next day, and output a prediction result according to the information on whether the hot migration corresponding to each preset time period can succeed.

[0118] Among them, it is possible to predict whether the hot migration of the virtual machine can succeed in multiple preset time periods in the next day based on Steps 301-306. For example, it can be determined that the hot migration of the virtual machine can succeed in each time period such as 0:00-1:00, 1:00-2:00...

[0119] Specifically, the electronic device can output a prediction result according to the information on whether the hot migration corresponding to each preset time period can succeed. For example, it can output the preset time periods for which the hot migration of the virtual machine is predicted to succeed, and use them as the recommended time periods for the virtual machine. It can also output the hot migration prediction results corresponding to each preset time period, so as to provide a reference basis for the hot migration of the virtual machine.

[0120] Figure 4 Schematic diagram of the prediction result of live migration of virtual machines shown in an exemplary embodiment of the present disclosure.

[0121] As Figure 4 shown, the prediction result indicates that live migration of virtual machines can succeed during several preset time periods such as 00:00 - 01:00, 01:00 - 02:00, 02:00 - 03:00, 03:00 - 04:00, 23:00 - 24:00 on a certain day in the future, while live migration of virtual machines will fail during other preset time periods.

[0122] Optionally, in any of the above implementation manners, the success rate during multiple preset time periods on the day after the current date can be predicted based on the historical performance data of the previous N days before the current date. Using more recent historical performance data for prediction can obtain a more accurate prediction result.

[0123] Figure 5 Schematic diagram of the solution for determining the success rate of live migration shown in an exemplary embodiment of the present disclosure.

[0124] As Figure 5 shown, in the solution provided by the present disclosure, for example, if the current date is the Tth day, the historical performance data from the (T - 1)th day to the (T - 8)th day can be obtained. Furthermore, the characteristic information for each preset time period during the time from the (T - 1)th day to the (T - 8)th day can be obtained. For example, the characteristic information for the time period of 00:00 - 01:00, and again for example, the characteristic information for the time period of 01:00 - 02:00…, the characteristic information for the time period of 23:00 - 24:00.

[0125] The characteristic information for each time period can be input into a preset prediction model, and then the success rate of live migration of virtual machines for each time period on the (T + 1)th day can be obtained. Furthermore, the success rate for each time period can be compared with a preset threshold to obtain the result of whether live migration of virtual machines for each time period is successful.

[0126] In an alternative exemplary embodiment, the present disclosure also provides a method for live migration of virtual machines, specifically including:

[0127] According to Figures 2 - 4 any of the above methods, determine the success rate when live migrating the target virtual machine during a preset time period in the future, and determine the migration time according to the success rate.

[0128] Live migrate the target virtual machine at the migration time.

[0129] Among them, the success rate of live migrating the target virtual machine during different time periods can be determined based on the historical performance data and static attribute information of the target virtual machine according to any of the above methods.

[0130] If it is predicted that the success rate of hot migrating the target virtual machine is relatively high within a preset time period in the future, any migration time can be selected within this preset time period. When this migration time is reached, the target virtual machine can be automatically hot migrated.

[0131] In an alternative implementation, if based on Figure 2 the shown solution, the success rates of hot migrating the target virtual machine within multiple preset time periods in a day are determined, then the migration time can be selected within the preset time period with the highest success rate.

[0132] In another alternative implementation, if based on Figure 3 the shown solution, the success rates of hot migrating the target virtual machine within multiple preset time periods in a day are determined, and at least one successful time period with a success rate greater than a preset threshold is determined among the multiple preset time periods, then the migration time can be selected within these successful time periods.

[0133] In this way, the success rate of hot migrating the target virtual machine can be improved, and the hot migration timing of the target virtual machine can be planned in advance before hot migration.

[0134] Figure 6 It is a schematic flowchart of a training method for a model for predicting the result of virtual machine hot migration shown in an exemplary embodiment of the present disclosure.

[0135] As Figure 6 shown, the training method for a model for predicting the result of virtual machine hot migration provided by the present disclosure includes:

[0136] Step 601, obtain the training performance data of the virtual machine within the preset time period every day in N days, the static attribute information of the virtual machine, and the migration result information of the virtual machine within the preset time period after N days.

[0137] Among them, the offline data of the virtual machine and the migration result information of migrating this virtual machine can be obtained, so that the training data for training the model can be obtained by using these data.

[0138] Specifically, the training performance data of the virtual machine within the preset time period every day in N days can be obtained from the offline data. The training performance data can include, for example, the data corresponding to each moment within the preset time period; the data corresponding to each moment can further include at least one of the following:

[0139] the usage rate of the central processing unit (CPU) of the virtual machine, the memory usage rate of the virtual machine, the network traffic data of the virtual machine, the information on the number of data read and write times of the virtual machine, the memory bandwidth data of the virtual machine.

[0140] Furthermore, the static attribute information of the virtual machine can include:

[0141] The first attribute information of the virtual machine itself and the second attribute information of the physical machine where the virtual machine is located.

[0142] The first attribute information includes at least one of the following: the number of central processing units of the virtual machine, the memory size of the virtual machine, and the operating system information of the virtual machine;

[0143] The second attribute information includes at least one of the following: the number of cores of the central processing unit of the physical machine, the model information of the physical machine, and the network card information of the physical machine.

[0144] For example, a prediction date can be set, and the training performance data within a preset time period for each day in the N days before the prediction date can be obtained. The migration result information within the preset time period after N days can be the migration result on the second day of the prediction date. For example, migration success or migration failure.

[0145] Step 602: Determine the training sample information of the virtual machine within a preset time period according to the training performance data and the static attribute information.

[0146] Among them, the training performance characteristics of the virtual machine within the preset time period can be determined according to the data corresponding to each moment within the preset time period in the training performance data;

[0147] Concatenate the training performance characteristics of the virtual machine with the static attribute information of the virtual machine to obtain the training sample information of the virtual machine within the preset time period.

[0148] Determining the training performance characteristics includes at least one of the following:

[0149] Determine the CPU performance characteristics according to the CPU usage rate of the virtual machine corresponding to each moment within the preset time period; the CPU performance characteristics include at least one of the following: the average value of the CPU usage rate, the quantile value of the CPU usage rate, and the variance of the CPU usage rate;

[0150] Determine the first memory performance characteristics according to the memory usage rate of the virtual machine corresponding to each moment within the preset time period; the first memory performance characteristics include at least one of the following: the average value of the memory usage rate, the quantile value of the memory usage rate, and the variance of the memory usage rate;

[0151] Determine the network performance characteristics according to the network traffic data of the virtual machine corresponding to each moment within the preset time period; the network performance characteristics include at least one of the following: the average value of the network traffic, the quantile value of the network traffic, and the variance of the network traffic;

[0152] Determine the data read / write performance characteristics according to the data read / write times information of the virtual machine corresponding to each moment within the preset time period; the data read / write performance characteristics include at least one of the following: the average value of data read / write times, the quantile value of data read / write times, and the variance of data read / write times.

[0153] Determine the second memory performance characteristics according to the memory bandwidth data of the virtual machine corresponding to each moment within the preset time period; the second memory performance characteristics include at least one of the following: the average value of memory bandwidth, the quantile value of memory bandwidth, and the variance of memory bandwidth.

[0154] Step 603: Input the training sample information into a preset model to obtain the success rate when hot migrating a virtual machine within a preset time period N days later.

[0155] The determined training sample information can be input into a preset model, which can be, for example, a neural network model, a binary classification model, etc.

[0156] Among them, the training sample information can be input into a preset model, and the preset model can output the success rate when hot migrating a virtual machine within a preset time period N days later.

[0157] For example, the training sample information within a preset time period can be determined using the training performance data N days before the prediction date, and inputting this training sample information into the preset model can obtain the success rate when hot migrating a virtual machine within a preset time period on the day after the prediction date.

[0158] Step 604: Adjust the parameters in the preset model according to the success rate and the migration result information of the virtual machine within a preset time period N days later, where the preset model after meeting the stop training condition is the model for predicting the hot migration result of the virtual machine.

[0159] Specifically, the success rate can be compared with a preset threshold to determine the result of whether the preset model can successfully hot migrate a virtual machine within a preset time period N days later, and it can also be compared with the migration result information of the virtual machine within a preset time period N days later obtained based on offline data, and then adjust the parameters in the preset model based on the comparison result until the stop training condition is met.

[0160] Furthermore, the stop training condition can be, for example, reaching a preset number of training times, or for another example, the difference between the prediction result output by the preset model and the migration result information of the virtual machine within a preset time period N days later is relatively small.

[0161] In actual application, a loss function can be constructed according to the result of whether the hot migration of the virtual machine can succeed predicted by the determined preset model after N days and the migration result information of the virtual machine in the preset time period after N days, and then the parameters in the preset model are adjusted by using the loss function, so that a model for predicting the hot migration result of the virtual machine can be obtained through multiple iterations.

[0162] Figure 7 The structure diagram of the virtual machine hot migration result prediction device shown in an exemplary embodiment of the present application.

[0163] As Figure 7 shown, the virtual machine hot migration result prediction device 700 provided in this embodiment includes:

[0164] An acquisition unit 710, configured to acquire the historical performance data of the virtual machine in a preset time period every day in the past N days and the static attribute information of the virtual machine; N is an integer greater than or equal to 1;

[0165] A feature determination unit 720, configured to determine the feature information of the virtual machine in the preset time period according to the historical performance data and the static attribute information;

[0166] A prediction unit 730, configured to input the feature information into a preset prediction model to obtain the success rate of hot migrating the virtual machine in a future preset time period, where the success rate is used to characterize the predicted migration result of the virtual machine.

[0167] In the solution provided by the present disclosure, the success rate of hot migrating the virtual machine in a future preset time period can be predicted by using the historical data of the virtual machine in the preset time period. This method not only considers the static characteristics of the virtual machine, but also considers the dynamic characteristics of the virtual machine in different preset time periods. Therefore, the solution provided by the present disclosure can leave enough time for the hot migration planning of the virtual machine.

[0168] Figure 8 The structure diagram of the virtual machine hot migration result prediction device shown in another exemplary embodiment of the present application.

[0169] As Figure 8 shown, in the virtual machine hot migration result prediction device 800 provided in this embodiment, on the basis of the device shown in Figure 7 the historical performance data includes data corresponding to each moment in the preset time period;

[0170] The feature determination unit 720 includes:

[0171] A performance characteristic determination module 721, configured to determine historical performance characteristics of the virtual machine within the preset time period according to data corresponding to each moment within the preset time period in the historical performance data;

[0172] A feature information determination module 722, configured to splice the historical performance characteristics of the virtual machine and the static attribute information of the virtual machine to obtain feature information of the virtual machine within the preset time period.

[0173] Optionally, data corresponding to any moment within the preset time period includes at least one of the following:

[0174] The CPU usage rate of the virtual machine, the memory usage rate of the virtual machine, the network traffic data of the virtual machine, the data read / write times information of the virtual machine, and the memory bandwidth data of the virtual machine.

[0175] Optionally, the performance characteristic determination module 721 is configured to perform at least one of the following steps:

[0176] Determine CPU performance characteristics according to the CPU usage rate of the virtual machine corresponding to each moment within the preset time period; the CPU performance characteristics include at least one of the following: the average value of the CPU usage rate, the quantile value of the CPU usage rate, and the variance of the CPU usage rate;

[0177] Determine first memory performance characteristics according to the memory usage rate of the virtual machine corresponding to each moment within the preset time period; the first memory performance characteristics include at least one of the following: the average value of the memory usage rate, the quantile value of the memory usage rate, and the variance of the memory usage rate;

[0178] Determine network performance characteristics according to the network traffic data of the virtual machine corresponding to each moment within the preset time period; the network performance characteristics include at least one of the following: the average value of the network traffic, the quantile value of the network traffic, and the variance of the network traffic;

[0179] Determine data read / write performance characteristics according to the data read / write times information of the virtual machine corresponding to each moment within the preset time period; the data read / write performance characteristics include at least one of the following: the average value of the data read / write times, the quantile value of the data read / write times, and the variance of the data read / write times;

[0180] Determine second memory performance characteristics according to the memory bandwidth data of the virtual machine corresponding to each moment within the preset time period; the second memory performance characteristics include at least one of the following: the average value of the memory bandwidth, the quantile value of the memory bandwidth, and the variance of the memory bandwidth.

[0181] Optionally, the obtaining unit 710 is specifically configured to:

[0182] Obtain the first attribute information of the virtual machine itself and the second attribute information of the physical machine where the virtual machine is located.

[0183] Optionally, the first attribute information includes at least one of the following: the number of central processing units of the virtual machine, the memory size of the virtual machine, and the operating system information of the virtual machine;

[0184] The second attribute information includes at least one of the following: the number of cores of the central processing unit of the physical machine, the model information of the physical machine, and the network card information of the physical machine.

[0185] Optionally, the device further includes:

[0186] A comparison unit 740, configured to compare the success rate with a preset threshold, and determine whether the hot migration of the virtual machine can succeed in a preset time period in the future according to the comparison result.

[0187] Optionally, a day includes multiple preset time periods; the device further includes an output unit 750, configured to:

[0188] Determine whether the hot migration of the virtual machine can succeed in multiple preset time periods in the future one day, and output a prediction result according to the information on whether the hot migration corresponding to each preset time period can succeed.

[0189] In an alternative embodiment, the present disclosure further provides a hot migration device for a virtual machine, specifically including:

[0190] A migration time determination unit, configured to determine, according to Figure 7 or Figure 8 the virtual machine hot migration result prediction device shown, the success rate when hot migrating a target virtual machine in a preset time period in the future, and determine the migration time according to the success rate;

[0191] A hot migration unit, configured to hot migrate the target virtual machine at the migration time.

[0192] Figure 9 It is a structural diagram of a training device for a model for predicting the hot migration result of a virtual machine shown in an exemplary embodiment of the present application.

[0193] As Figure 9 shown, the training device 900 for a model for predicting the hot migration result of a virtual machine provided in this embodiment includes:

[0194] An acquisition unit 910, configured to acquire the training performance data of the virtual machine in the preset time period every day in N days, the static attribute information of the virtual machine, and the migration result information of the virtual machine in the preset time period after N days;

[0195] A sample determination unit 920, configured to determine training sample information of the virtual machine in the preset time period according to the training performance data and the static attribute information;

[0196] A prediction unit 930, configured to input the training sample information into a preset model to obtain a success rate when the virtual machine is thermally migrated in a preset time period N days later;

[0197] An adjustment unit 940, configured to adjust parameters in the preset model according to the success rate and migration result information of the virtual machine in the preset time period N days later, where the preset model after meeting the training stop condition is a model for predicting the result of virtual machine thermal migration.

[0198] Figure 10 It is a structural diagram of an electronic device shown in an exemplary embodiment of the present application.

[0199] As Figure 10 shown, the electronic device provided in this embodiment includes:

[0200] A memory 101;

[0201] A processor 102; and

[0202] A computer program;

[0203] Wherein, the computer program is stored in the memory 101 and is configured to be executed by the processor 102 to implement any one of the above-mentioned virtual machine thermal migration result prediction methods or the training method of the model for predicting the virtual machine thermal migration result.

[0204] This embodiment also provides a computer-readable storage medium, on which a computer program is stored,

[0205] The computer program is executed by a processor to implement any one of the above-mentioned virtual machine thermal migration result prediction methods or the training method of the model for predicting the virtual machine thermal migration result.

[0206] This embodiment also provides a computer program, including program code. When a computer runs the computer program, the program code executes any one of the above-mentioned virtual machine thermal migration result prediction methods or the training method of the model for predicting the virtual machine thermal migration result.

[0207] Those of ordinary skill in the art can understand that all or part of the steps of implementing the above method embodiments can be completed by hardware related to program instructions. The foregoing program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps including the above method embodiments; and the foregoing storage medium includes: various media such as ROM, RAM, magnetic disk, or optical disk that can store program code.

[0208] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for predicting the result of live migration of virtual machines, characterized in that, Including: Obtaining historical performance data of a virtual machine within a preset time period every day in the past N days and static attribute information of the virtual machine; N is an integer greater than or equal to 1; the historical performance data includes data corresponding to each moment within the preset time period; Determining historical performance characteristics of the virtual machine within the preset time period according to the data corresponding to each moment within the preset time period in the historical performance data; Concatenating the historical performance characteristics of the virtual machine and the static attribute information to obtain characteristic information of the virtual machine within the preset time period; Inputting the characteristic information into a preset prediction model to obtain a success rate when live migrating the virtual machine in a future preset time period, where the success rate is used to characterize the predicted migration result of the virtual machine; The determining historical performance characteristics of the virtual machine within the preset time period according to the data corresponding to each moment within the preset time period in the historical performance data includes at least one of the following: Determining CPU performance characteristics according to the CPU usage rate of the virtual machine corresponding to each moment within the preset time period; the CPU performance characteristics include at least one of the following: average CPU usage rate, CPU usage rate quantile value, CPU usage rate variance; wherein, the historical performance data includes the CPU usage rate of the virtual machine; Determining first memory performance characteristics according to the memory usage rate of the virtual machine corresponding to each moment within the preset time period; the first memory performance characteristics include at least one of the following: average memory usage rate, memory usage rate quantile value, memory usage rate variance; wherein, the historical performance data includes the memory usage rate of the virtual machine; Determining network performance characteristics according to the network traffic data of the virtual machine corresponding to each moment within the preset time period; the network performance characteristics include at least one of the following: average network traffic, network traffic quantile value, network traffic variance; wherein, the historical performance data includes the network traffic data of the virtual machine; Determining data read / write performance characteristics according to the data read / write times information of the virtual machine corresponding to each moment within the preset time period; the data read / write performance characteristics include at least one of the following: average data read / write times, data read / write times quantile value, data read / write times variance; wherein, the historical performance data includes the data read / write times information of the virtual machine; Determining second memory performance characteristics according to the memory bandwidth data of the virtual machine corresponding to each moment within the preset time period; the second memory performance characteristics include at least one of the following: average memory bandwidth, memory bandwidth quantile value, memory bandwidth variance; wherein, the historical performance data includes the memory bandwidth data of the virtual machine.

2. The method according to claim 1, characterized in that, Obtaining the static attribute information of the virtual machine includes: Obtaining first attribute information of the virtual machine itself and second attribute information of the physical machine where the virtual machine is located.

3. The method according to claim 2, characterized in that, The first attribute information includes at least one of the following: the number of central processing units of the virtual machine, the memory size of the virtual machine, the operating system information of the virtual machine; The second attribute information includes at least one of the following: the number of cores of the central processing unit of the physical machine, the model information of the physical machine, and the network card information of the physical machine.

4. The method according to any one of claims 1 - 3, characterized in that, It further includes: Compare the success rate with a preset threshold, and determine whether the hot migration of the virtual machine can succeed within a preset time period in the future according to the comparison result.

5. The method according to claim 4, characterized in that, A day duration includes multiple preset time periods; the method further includes: Determine whether the hot migration of the virtual machine can succeed within multiple preset time periods in the next day, and output a prediction result according to the information on whether the hot migration corresponding to each preset time period can succeed.

6. A live migration method for virtual machines, characterized in that, It includes: According to the method according to any one of claims 1-5, determine the success rate when hot migrating a target virtual machine within a preset time period in the future, and determine the migration time according to the success rate. Hot migrate the target virtual machine at the migration time.

7. A training method for a model used to predict the result of live migration of virtual machines, characterized in that, It includes: Obtain the training performance data of the virtual machine within the preset time period every day in N days, the static attribute information of the virtual machine, and the migration result information of the virtual machine within the preset time period after N days. According to the data corresponding to each moment within the preset time period in the training performance data, determine the training performance characteristics of the virtual machine within the preset time period. Concatenate the training performance characteristics of the virtual machine with the static attribute information to obtain the training sample information of the virtual machine within the preset time period. Input the training sample information into a preset model to obtain the success rate when hot migrating the virtual machine within the preset time period after N days. Adjust the parameters in the preset model according to the success rate and the migration result information of the virtual machine within the preset time period after N days, where the preset model after meeting the stop training condition is a model for predicting the hot migration result of the virtual machine. The determination of the training performance characteristics of the virtual machine within the preset time period includes at least one of the following: According to the CPU usage rate of the virtual machine corresponding to each moment within the preset time period, determine the CPU performance characteristics; the CPU performance characteristics include at least one of the following: the average value of the CPU usage rate, the quantile value of the CPU usage rate, and the variance of the CPU usage rate. According to the memory usage rate of the virtual machine corresponding to each moment within the preset time period, determine the first memory performance characteristics; the first memory performance characteristics include at least one of the following: the average value of the memory usage rate, the quantile value of the memory usage rate, and the variance of the memory usage rate. According to the network traffic data of the virtual machine corresponding to each moment within the preset time period, determine the network performance characteristics; the network performance characteristics include at least one of the following: the average value of the network traffic, the quantile value of the network traffic, and the variance of the network traffic. According to the data read / write times information of the virtual machine corresponding to each moment within the preset time period, determine the data read / write performance characteristics; the data read / write performance characteristics include at least one of the following: the average value of the data read / write times, the quantile value of the data read / write times, and the variance of the data read / write times. Determine a second memory performance characteristic according to the memory bandwidth data of the virtual machine corresponding to each moment within the preset time period; the second memory performance characteristic includes at least one of the following: mean memory bandwidth, memory bandwidth quantile value, memory bandwidth variance.

8. A virtual machine live migration result prediction device, characterized in that Comprising: An acquisition unit, configured to acquire historical performance data of the virtual machine within a preset time period every day in the past N days and static attribute information of the virtual machine; N is an integer greater than or equal to 1; the historical performance data includes data corresponding to each moment within the preset time period; A feature determination unit, configured to determine a historical performance characteristic of the virtual machine within the preset time period according to the data corresponding to each moment within the preset time period in the historical performance data; Concatenate the historical performance characteristic of the virtual machine and the static attribute information to obtain characteristic information of the virtual machine within the preset time period; A prediction unit, configured to input the characteristic information into a preset prediction model to obtain a success rate when hot migrating the virtual machine in a future preset time period, where the success rate is used to characterize the predicted migration result of the virtual machine; The determining the historical performance characteristic of the virtual machine within the preset time period according to the data corresponding to each moment within the preset time period in the historical performance data includes at least one of the following: Determine a CPU performance characteristic according to the CPU usage rate of the virtual machine corresponding to each moment within the preset time period; the CPU performance characteristic includes at least one of the following: mean CPU usage rate, CPU usage rate quantile value, CPU usage rate variance; where the historical performance data includes the CPU usage rate of the virtual machine; Determine a first memory performance characteristic according to the memory usage rate of the virtual machine corresponding to each moment within the preset time period; the first memory performance characteristic includes at least one of the following: mean memory usage rate, memory usage rate quantile value, memory usage rate variance; where the historical performance data includes the memory usage rate of the virtual machine; Determine a network performance characteristic according to the network traffic data of the virtual machine corresponding to each moment within the preset time period; the network performance characteristic includes at least one of the following: mean network traffic, network traffic quantile value, network traffic variance; where the historical performance data includes the network traffic data of the virtual machine; Determine a data read / write performance characteristic according to the data read / write times information of the virtual machine corresponding to each moment within the preset time period; the data read / write performance characteristic includes at least one of the following: mean data read / write times, data read / write times quantile value, data read / write times variance; where the historical performance data includes the data read / write times information of the virtual machine; Determine a second memory performance characteristic according to the memory bandwidth data of the virtual machine corresponding to each moment within the preset time period; the second memory performance characteristic includes at least one of the following: mean memory bandwidth, memory bandwidth quantile value, memory bandwidth variance; where the historical performance data includes the memory bandwidth data of the virtual machine.

9. A virtual machine live migration device, characterized in that Comprising: A migration time determination unit, configured to determine, according to the device described in claim 8, the success rate of hot migrating a target virtual machine in a preset time period in the future, and determine the migration time according to the success rate; A hot migration unit, configured to hot migrate the target virtual machine at the migration time.

10. An electronic device, characterized in that Comprising: A memory; A processor; And A computer program; Wherein, the computer program is stored in the memory and configured to be executed by the processor to implement the method described in any one of claims 1-7.

11. A computer-readable storage medium, characterized in that On which a computer program is stored, The computer program is executed by a processor to implement the method described in any one of claims 1-7.

12. A computer program product, characterized in that Including a computer program, The computer program is executed by a processor to implement the method described in any one of claims 1-7.

Images