CN111381928B

CN111381928B - Virtual machine migration method, cloud computing management platform and storage medium

Info

Publication number: CN111381928B
Application number: CN201811628466.0A
Authority: CN
Inventors: 童遥; 孔鹏; 李华; 申光
Original assignee: ZTE Corp
Current assignee: ZTE Corp
Priority date: 2018-12-28
Filing date: 2018-12-28
Publication date: 2021-05-25
Anticipated expiration: 2038-12-28
Also published as: WO2020134364A1; CN111381928A

Abstract

The embodiment of the invention discloses a virtual machine migration method, which comprises the following steps: if the first load of the first physical machine at the current first moment is larger than a first threshold value, predicting a second load of the first physical machine at a second moment; wherein the second time is after the first time; if the second load is larger than a second threshold value, selecting a virtual machine to be migrated from the virtual machines on the first physical machine; wherein the second threshold is greater than the first threshold; determining a second physical machine, and migrating the virtual machine to be migrated to the second physical machine; wherein the second physical machine is different from the first physical machine. The embodiment of the invention also discloses a cloud computing management platform and a storage medium.

Description

Virtual machine migration method, cloud computing management platform and storage medium

Technical Field

The embodiment of the invention relates to the field of virtual machine resource scheduling of a cloud computing data center, in particular to a virtual machine migration method, a cloud computing management platform and a storage medium.

Background

The cloud computing management platform is a complex large-scale system, the virtual machines are an important part of the cloud computing management platform, and the cloud computing management platform has the characteristics of cross-system, resource isolation, migration and the like, and the efficient virtual machine cluster deployment and dynamic migration operation can integrate the dispersed virtual machines, meet the large-scale application requirements, improve the operation maintenance and service quality of the cloud computing management platform, and is an important component of the cloud computing management platform.

However, the cloud computing management platform has the characteristics of large resource scale, complex structure and wide region distribution, and the application of the cloud computing management platform has the characteristics of flexible and changeable requests, different resource type changes, dynamic load changes and the like, so that great challenges are provided for the cluster deployment and dynamic migration strategies of the virtual machines. The dynamic migration is caused by the requirement of load balancing, and under the cloud computing environment, the situation that some computers are overloaded and always in an overload state, and other computers are often in an idle state may occur, so that the overall resource utilization rate of the system is greatly reduced. Load balancing is the adjustment of tasks among multiple physical resources to achieve optimal overall utilization of resources.

The migration triggering strategy in the related art is based on a fixed threshold, and when the performance load of a certain type of a certain node is greater than the fixed threshold, a system triggers one migration; however, due to the existence of various types of bursty loads, once the bursty load at any moment reaches a fixed threshold, virtual machine migration is triggered, which results in frequent virtual machine migration, causes unnecessary migration overhead, and also results in system reliability being not guaranteed.

Disclosure of Invention

In view of this, embodiments of the present invention desirably provide a virtual machine migration method, a cloud computing management platform, and a storage medium, so as to solve the problem in the related art that frequent virtual machine migration is caused and unnecessary migration overhead is caused because instantaneous bursty load reaches a fixed threshold, avoid interference of the bursty load on a system, and improve reliability of the system.

In order to achieve the purpose, the technical scheme of the invention is realized as follows:

a virtual machine migration method is applied to a cloud computing management platform and comprises the following steps:

if the first load of the first physical machine at the current first moment is larger than a first threshold value, predicting the second load of the first physical machine at a second moment; wherein the second time is after the first time;

if the second load is larger than a second threshold value, selecting a virtual machine to be migrated from the virtual machines on the first physical machine; wherein the second threshold is greater than the first threshold;

determining a second physical machine, and migrating the virtual machine to be migrated to the second physical machine; wherein the second physical machine is different from the first physical machine.

Optionally, if the first load of the first physical machine at the current first time is greater than the first threshold, predicting the second load of the first physical machine at the second time includes:

if the first load is larger than the first threshold, acquiring third loads of at least two moments in a preset time period; the preset time period comprises the first moment and at least one third moment before the first moment, and the third load at each moment of the at least two moments comprises at least two loads with different attributes; the loads with different attributes represent different types of occupied resources in the first physical machine;

calculating a fourth load at each moment according to the third load at each moment and the weight value of the third load at each moment;

and calculating the second load according to the fourth load at each moment.

Optionally, the calculating the second load according to the fourth load at each time includes:

and calculating the average value of the fourth loads at a plurality of moments to obtain the second load.

Optionally, before obtaining the multiple different types of load data at each of multiple moments before the first moment if the first load is greater than the first threshold, the method further includes:

receiving a user request and acquiring the request type of the user request; wherein the request type comprises a static request or a dynamic request;

correspondingly, before the calculating the fourth load at each moment according to the third load at each moment and the weighted value of the third load at each moment, the method further includes:

acquiring the attribute of each load in the third loads at each moment;

and determining a weight value of each load in the third loads at each moment according to the request type and the attribute of each load in the third loads at each moment.

Optionally, the third load at each time further includes a distance, where the distance represents a distance between the first physical machine and a data center that receives a user request, and the cloud computing management platform further includes the data center, where the data center is configured to manage a third physical machine, and the third physical machine includes the first physical machine.

Optionally, before the calculating the fourth load at each time according to the third load at each time and the weighted value of the third load at each time, the method further includes:

acquiring the attribute of each load in the third loads at each moment;

and determining a weight value of each load in the third loads at each moment according to the attribute of each load in the third loads at each moment.

Optionally, the number of the virtual machines on the first physical machine is multiple, and if the second load is greater than a second threshold, selecting a virtual machine to be migrated from the virtual machines on the first physical machine includes:

if the second load is larger than the second threshold, acquiring the resource utilization rate of the first physical machine;

and selecting a virtual machine to be migrated from the plurality of virtual machines on the first physical machine according to the resource utilization rate.

Optionally, the selecting, according to the resource utilization, a virtual machine to be migrated from the multiple virtual machines on the first physical machine includes:

if the resource utilization rate is greater than or equal to a third threshold, predicting a fifth load of each virtual machine on the first physical machine at a third moment; wherein the third time is after the first time;

and selecting a virtual machine corresponding to the maximum load in the fifth load from the virtual machines on the first physical machine, and determining the virtual machine as the virtual machine to be migrated.

if the resource utilization rate is less than or equal to a fourth threshold value, determining each virtual machine on the first physical machine as the virtual machine to be migrated; wherein the fourth threshold is less than the third threshold.

Optionally, the determining the second physical machine includes:

acquiring a sixth load of each physical machine in the cloud computing management platform at the first moment;

selecting a physical machine with a sixth load smaller than a fifth threshold value from the physical machines in the cloud computing management platform, and determining the physical machine as a third physical machine; the third physical machine is different from the first physical machine, and the number of the third physical machines is multiple;

predicting a seventh load of virtual machines on each of the third physical machines at a fourth time; wherein the fourth time is after the first time;

calculating the sum of the seventh load of each third physical machine and the load of the virtual machine to be migrated to obtain a plurality of target loads;

determining a target load of the plurality of target loads that is less than a sixth threshold;

selecting a third physical machine corresponding to the target load smaller than a sixth threshold value from the third physical machines, and determining the third physical machine as a fourth physical machine; wherein the number of the fourth physical machines is at least one;

determining the second physical machine from at least one of the fourth physical machines.

Optionally, the determining the second physical machine from at least one fourth physical machine includes:

calculating a value obtained by subtracting the sixth load of each fourth physical machine from the seventh load of each fourth physical machine to obtain the resource demand expansion amount of each fourth physical machine;

calculating the total resource amount of each fourth physical machine minus the value of the sixth load of each fourth physical machine to obtain the unallocated resource amount of each fourth physical machine;

calculating a value obtained by subtracting the resource demand expansion amount of each fourth physical machine from the unallocated resource amount of each fourth physical machine to obtain a predicted residual resource amount of each fourth physical machine;

and determining the second physical machine from the fourth physical machines according to the predicted residual resource amount of each fourth physical machine.

Optionally, the determining the second physical machine from the fourth physical machine according to the predicted remaining resource amount of the fourth physical machine includes:

and determining the fourth physical machine corresponding to the maximum value in the predicted residual resource amount of the fourth physical machine as the second physical machine.

Optionally, a virtual machine cluster is established on the physical machines of the cloud computing management platform, the number of the virtual machines to be migrated is at least two, and the number of the virtual machines to be migrated is the same as that of the second physical machines; before migrating the virtual machine to be migrated to the second physical machine, the method further includes:

acquiring the load of each virtual machine to be migrated;

determining a target physical machine corresponding to each virtual machine to be migrated from a plurality of second physical machines; wherein the plurality of second physical machines includes the target physical machine;

acquiring loads of a plurality of target physical machines;

and sequencing the plurality of target physical machines according to the load of each virtual machine to be migrated and the load of each target physical machine to obtain a priority queue.

Optionally, the sorting the plurality of second physical machines according to the load of each to-be-migrated virtual machine and the load of each target physical machine to obtain a priority queue includes:

calculating the sum of the load of each virtual machine to be migrated and the load of a target physical machine corresponding to each virtual machine to be migrated to obtain a transmission load;

and sequencing the target physical machines according to the transmission load to obtain a priority queue.

Optionally, the migrating the virtual machine to be migrated to the second physical machine includes:

determining the priority of each target physical machine according to the priority queue;

carrying out mirror image transmission on the virtual machines to be migrated corresponding to each target physical machine according to the sequence of the priority from high to low; and the transmission load corresponding to the target physical machine with the high priority is smaller than the transmission load corresponding to the target physical machine with the low priority.

A cloud computing management platform, the cloud computing management platform comprising:

a memory for storing executable instructions;

and the processor is used for executing the executable instructions stored in the memory to realize the steps in the virtual machine migration method.

A storage medium storing one or more programs, the one or more programs being executable by one or more processors to implement the steps of the virtual machine migration method as described above.

According to the virtual machine migration method, the cloud computing management platform and the storage medium provided by the embodiment of the invention, if the first load of the first physical machine at the current first moment is greater than the first threshold, the second load of the first physical machine at the second moment is predicted; wherein the second time is after the first time; if the second load is larger than a second threshold value, selecting a virtual machine to be migrated from the virtual machines on the first physical machine; wherein the second threshold is greater than the first threshold; determining a second physical machine, and migrating the virtual machine to be migrated to the second physical machine; wherein the second physical machine is different from the first physical machine; that is to say, in the embodiment of the present invention, when it is determined that the bursty load of the first physical machine at the first time is greater than the first threshold, the load of the first physical machine at a future time is predicted, and when it is determined that the load at the future time is higher than the second threshold that is greater than the first threshold, a migration is triggered, so that interference of the bursty load on the system is effectively avoided, the problem that the virtual machine migration is triggered when the instantaneous bursty load reaches the fixed threshold in the related art, so that frequent virtual machine migration is caused, and unnecessary migration overhead is caused is solved, interference of the bursty load on the system is avoided, and reliability of the system is improved.

Drawings

Fig. 1 is a schematic functional architecture diagram of a cloud computing platform management system according to an embodiment of the present invention;

fig. 2 is a schematic diagram of an implementation architecture of a cloud computing platform management system according to an embodiment of the present invention;

fig. 3 is a schematic diagram of an agent implementation architecture according to an embodiment of the present invention;

fig. 4 is a schematic diagram of a virtual machine migration model according to an embodiment of the present invention;

fig. 5 is a schematic flowchart of a virtual machine migration method according to an embodiment of the present invention;

fig. 6 is a schematic flowchart of another virtual machine migration method according to an embodiment of the present invention;

fig. 7 is a schematic flowchart of another virtual machine migration method according to an embodiment of the present invention;

fig. 8 is a schematic diagram of a virtual machine cluster deployment architecture based on resource classification according to an embodiment of the present invention;

fig. 9 is a flowchart illustrating a virtual machine migration method according to another embodiment of the present invention;

fig. 10 is a flowchart illustrating another virtual machine migration method according to another embodiment of the present invention;

fig. 11 is a schematic structural diagram of a cloud computing platform management system according to an embodiment of the present invention.

Detailed Description

The technical solution in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

It should be appreciated that reference throughout this specification to "an embodiment of the present invention" or "an embodiment described previously" means that a particular feature, structure or characteristic described in connection with the embodiment is included in at least one embodiment of the present invention. Thus, the appearances of the phrase "in an embodiment of the present invention" or "in the foregoing embodiments" in various places throughout the specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In various embodiments of the present invention, the sequence numbers of the above-mentioned processes do not mean the execution sequence, and the execution sequence of each process should be determined by its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention. The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.

Before further detailed description of the embodiments of the present invention, a virtual machine migration technique in the related art is described.

However, the cloud computing management platform has the characteristics of large resource scale, complex structure and wide region distribution, and the application of the cloud computing management platform has the characteristics of flexible and changeable requests, different resource type changes, dynamic load changes and the like, so that great challenges are provided for the cluster deployment and dynamic migration strategies of the virtual machines. The dynamic migration occurs because of the requirement for load balancing, and under the cloud computing environment, the situation that some computers are overloaded and always in an overload state, and other computers are often in an idle state may occur, so that the overall resource utilization rate of the system is greatly reduced. Load balancing is the adjustment of tasks among multiple physical resources to achieve optimal overall utilization of resources.

According to the characteristics of cloud computing management platform resources, the main difficulties of dynamic migration can be known through analysis in the following three aspects: handling of sudden load, huge number of system virtual machines, guarantee of system reliability and the like.

In the embodiment of the present invention, the above three aspects are briefly introduced:

first aspect, handling of sudden load

With the widespread use of cloud computing management platforms, the user size may become very large at a certain time, which is one of the bursty loads. Taking the visit volume of a certain website on 11 th day in 2017 and 11 th day as an example, the visit volume of a user for the website in a short time can reach ten million levels, which is far higher than the average value. Such bursty load behavior poses significant challenges for system performance analysis and live migration of virtual machines.

Second, the number of virtual machines of the system is huge

The cloud computing management platform has thousands of servers, and each physical machine may have tens of virtual machines, and the resource types of the virtual machines are different, and performance data is generated all the time. In the face of such large-scale servers and data, it is difficult to directly use the traditional method to perform performance analysis on the cloud computing management platform.

Third aspect, system reliability guarantee

The reliability of the system guarantees that on one hand, dynamic migration is required to be incapable of causing too much system overhead to interfere with normal operation of the system, and on the other hand, dynamic migration cannot cause interference to an application program deployed in a virtual machine. This requires that the live migration selects a virtual machine that actually needs to be migrated, and a physical node whose load will not be excessively increased in the future, so as to ensure that the virtual machine migration will not interfere with the existing virtual machine and application.

It should be noted that, in addition to the above three aspects, attention is paid to prevention of oscillation after the virtual machine is migrated. Because a plurality of nodes may select the same physical node machine as a target node based on current load information, the load of the selected target node is increased sharply, and secondary or even tertiary migration is caused after the load exceeds a threshold set by an administrator, that is, migration shock is generated.

In the related art, some systems adopt an operating system virtualization technology, for example, a non-downtime migration technology, and migrate a virtual machine to another server with a light load without the awareness of a user, which provides a completely new experience for the periodic management and maintenance and error recovery of the system. Although the system consumes less performance when migrating, the system cannot effectively apply bursty loads.

In the related technology, some systems do not need to modify an operating system and a kernel system during migration, so that services can run continuously; however, in the migration process, the migration entity is an operating system image and application, the migration efficiency is low, and the problem of migration oscillation is not solved.

In the related art, a seamless migration technology between the current server based on the intel processor and the future technology exists, the technology has good backward compatibility, but the migration system does not fully consider the performance relationship before and after migration, and may migrate shock and increase the system overhead.

As can be known from the virtual machine migration technology adopted in the related art, the existing dynamic migration of a virtual machine still has defects, specifically:

(1) the migration strategy of most load balancing methods is only based on a certain fixed threshold, that is, the load of a node machine exceeds the upper bound of the load threshold, the node machine triggers migration, so that an instantaneous burst load peak value triggers virtual machine migration, and unnecessary migration overhead is easily caused.

(2) The current load prediction method is used for predicting the load of a virtual machine, predicting the application load and partially predicting the load of a physical host, the index range of the load is narrow, a singular value algorithm, incremental weighting and the like are used in a prediction means, but the prediction aims to screen out a physical machine which meets the conditions and migrate all virtual machines on the physical machine, and the same physical node machine may be selected as a target node in the migration process, so that the load of the selected target node is increased sharply, and migration shock is caused.

(3) The method for cluster deployment related to virtual machine migration in the prior art includes the steps of sorting virtual machines and physical machines according to Central Processing Unit (CPU) occupancy rates or memory occupancy rates respectively, then matching the virtual machines and the physical machines one by one, or sorting the ratio by using the ratio of the CPU occupancy rates and the memory occupancy rates of the virtual machines, wherein sorting results are matched with sorting results of the physical machines, reference indexes are narrow, and improvement on virtual machine migration efficiency is not obvious.

(4) The existing mirror image transfer methods related to virtual machine migration are all based on the sequential copying of the cache of the virtual machine mirror image to a target physical host, and when the size of a virtual machine cluster to be migrated is large, the defect of low efficiency exists.

(5) The current virtual machine migration only considers the scenario of a single data center and migrates to the physical host closest to the data center, and does not consider the scenario of multiple data centers, that is, there is no means for preventing virtual machines from migrating across data centers.

According to the foregoing embodiments, an embodiment of the present invention provides a virtual machine migration method, which is applied to a cloud computing management platform, where the cloud computing management platform may also be referred to as a cloud computing management platform system, and may also be referred to as a system hereinafter; the system adopts a multi-level distributed architecture, the whole system is provided with a plurality of data centers, each data center is provided with a plurality of physical machines and a data center server, and the data centers are connected to the system server through a network.

The client accesses the system server, one system server manages all the data centers, each data center is provided with one data center server, the data center servers manage all the physical machines of the data center, and each physical machine is provided with an agent which is responsible for processing commands sent by the data center servers and responding to operation results. The distributed multi-layer management structure can conveniently realize the expansion of the system and is beneficial to relieving the pressure of the server.

Referring to fig. 1, the cloud computing management platform system 10 includes an image file management module 110, a cluster management module 120, a performance monitoring module 130, and a resource management module 140, which are described below.

The Image file management module (Image Manager)110 may upload and download images, classify resource types of the images, and manage copies, where the resource types of the images are initially specified by a user and may be modified by the system at a later stage.

The virtual machine Cluster management module (Cluster Manager)120 may implement editing of the Cluster, deployment of the Cluster, and deletion of the Cluster.

The Performance monitoring module (Performance Monitor)130 may be configured to Monitor the Performance of the physical machine, Monitor the Performance of the virtual machine, and predict the load Performance of the system.

The Resource management module (Resource Manager)140 may implement physical Resource addition, physical Resource deletion, physical Resource editing, and mirror image library management.

The cloud computing management platform system implementation architecture is shown in fig. 2, and includes a presentation layer 210, a logic layer 220, and a data persistence layer 230, which are respectively described below.

The presentation layer 210 is a display interface for interaction between a user and the system, and shows system performance, operation results, and cluster operation conditions through contents such as tables, graphics, and characters.

The logic layer 220, which is a core layer of the cluster management policy, is responsible for responding to the operation commands sent by the presentation layer and provides an access Interface to show the system operation status, and provides 5 sub-modules, which are respectively a Resource management (Resource Manager) module, a Performance management (Performance Manager) module, a Virtualization Engine (Virtualization Engine) module, an Agent management (Agent Manager) module, and a front-end interaction Interface API Interface module.

The Agent Manager module is mainly deployed on each physical machine and virtual machine of the system, and an Agent (Agent) on the physical machine is mainly used for responding to operation commands sent by a data center server, such as virtual machine creation, starting, stopping, destroying, migration and the like, and collecting performance data and submitting the performance data to the system. The Agent in the virtual machine is mainly used to send the current performance status of the virtual machine to the system, the collected performance data mainly includes CPU utilization, network bandwidth utilization, memory utilization and file system status within a period of time, in addition, operations such as transferring images, deploying files in a cluster, starting a cluster and the like all need to be involved in the Agent installed on each physical machine, the Agent also needs to be responsible for sending the performance state information of the machine to the server, the implementation architecture of the Agent is shown in fig. 3, and includes an image file processing module 310, a virtualization event processing module 320 and a performance monitoring module 330, which are described below.

The image processing module 310 may receive the image and transmit the image.

The virtualization event processing module 320 may implement virtual machine startup, virtual machine creation, virtual machine shutdown, and virtual machine deployment.

The performance monitoring module 330 may be configured to perform performance monitoring on the database program, the file transfer program, and the XML program.

The Performance Manager module is responsible for analyzing the data collected by the Agent Manager and determining the current operation condition of the whole and each physical. The Performance Manager module can predict the Performance condition of the system at the next time, and when the load is too heavy, the Performance Manager module can send a warning of the too heavy load to the system to remind a Manager to add physical resources or adjust the load through a load balancing means. The module provides decision support for deployment and migration of the virtual machine and for a user to check the current system operation condition.

The Virtualization Engine module is a core module of the system, and can receive a request of a user to implement a series of Virtualization operations, including events such as creation of a virtual machine, starting of the virtual machine, stopping of the virtual machine, destruction of the virtual machine and the like, after receiving a cluster deployment request sent from an upper layer, performing XML analysis on the cluster deployment request, and if the cluster deployment request is judged to be deployable, performing a cluster deployment process, such as transferring a mirror image, configuring a cluster deployment file, starting the cluster and the like, and simultaneously sending the operated Virtualization events to an Agent Manager module in a command form for specific operation, receiving information of the Performance Manager module, and migrating part of the virtual machines to lighter-loaded machines when the machines are overloaded.

The Resource Manager module is mainly responsible for managing physical resources and monitoring the current use condition of the physical resources, wherein the physical resources mainly comprise a machine CPU, a memory, a file system and a network. The module also manages the resources of each mirror bank.

The data persistence layer 230 provides various methods for data persistence, a Database program (Database Provider) provides support for accessing a Database, a File-Server Provider (File-Server Provider) persists large data in a File form, such as an image File, and an XML program (XML Provider) is mainly used for support for reading and writing an XML File.

In the embodiment of the present invention, the cloud computing management platform system 10 further includes a virtual machine dynamic migration controller (not shown in fig. 1).

The virtual machine dynamic migration device is composed of two large modules, namely a workload prediction module and a migration control module, which are shown in fig. 4, and the migration control module may include a migration trigger module, a virtual machine to be migrated selection module, and a target physical machine selection module; the migration triggering module is used for determining a migration time, namely determining when a virtual machine should be migrated so as to ensure that the performance of the whole physical machine is always kept in a better state; the virtual machine to be migrated selecting module is used for determining which virtual machines should be migrated; and the target physical machine selection module is used for determining the target physical machine to which the virtual machine to be migrated should be migrated.

Based on the foregoing embodiments, the embodiment of the present invention provides a virtual machine migration method, which is applied to a cloud computing management platform; as shown in fig. 5, the method comprises the steps of:

step 101, if the first load of the first physical machine at the current first moment is greater than a first threshold, predicting a second load of the first physical machine at a second moment.

Wherein the second time is after the first time.

In the embodiment of the invention, the cloud computing management platform determines that the first load of the first physical machine at the current first moment is greater than the first threshold, does not directly trigger the migration of the virtual machine on the first physical machine, and predicts the second load of the first physical machine at the second moment, namely the future moment, so as to further determine whether to trigger the migration of the virtual machine by means of the second load.

In the embodiment of the invention, due to the diversification of the application in the virtual machine, the balance degrees of events, resource use and load of the virtual machine are also greatly different, and different migration strategies can generate great difference on the migration cost. In order to support the decision-making system to select the correct virtual machine and migration destination to be migrated, it is necessary to predict a key indicator of workload.

In the embodiment of the invention, the system can acquire the load data of the physical machine in real time and can also acquire the load data of the physical machine periodically, thereby laying a foundation for predicting the load data of the physical machine. For example, in practical applications, a total of 288 sample points can be acquired within 24 hours with 5 minutes as a sampling period.

In the embodiment of the invention, after the load data of the physical machine is collected in real time, the load data is stored, and further, the CPU utilization rate data of various services on the physical machine can be used as the dynamic demand of the CPU parameters of the cloud task executed by the virtual machine of the physical machine.

In the embodiment of the present invention, the load at the second time T2 may be calculated from the load data at the first time T1 and the load data at a plurality of times within a specified time period before T1.

In another embodiment of the present invention, when it is determined that the first physical machine has the bursty load at time T1, the load at the second time T2 may be calculated by using load data at a plurality of times within a specified time period before T1.

Referring to fig. 6, a migration controller in the system completes a decision of migration, selection of a virtual machine to be migrated, and migration of the virtual machine to be migrated to a target physical machine under guidance of a control module in the migration controller through cooperation among a migration trigger module, a virtual machine to be migrated selection module, and a target physical machine selection module. In the embodiment of the invention, the virtual machine dynamic migration controller adopts a performance prediction algorithm to realize the determination of the migration opportunity and the selection of the migration target physical machine.

And 102, if the second load is larger than a second threshold value, selecting a virtual machine to be migrated from the virtual machines on the first physical machine.

Wherein the second threshold is greater than the first threshold.

In the embodiment of the invention, the cloud computing management platform triggers one migration only when determining that the second load is greater than the second threshold, namely determining that the load at the future moment is higher than the second threshold which is greater than the first threshold; and, once it is determined that virtual machine migration is triggered, selecting some or all of the virtual machines from the first physical machine for migration.

As can be seen from the above, in the embodiment of the present invention, through the migration triggering module, the system can determine the performance load condition of one physical machine node, and determine the time for triggering migration. The migration triggering policy in the related art is based on a fixed threshold, and when some type of performance load of a certain physical machine node exceeds the specified threshold, the system triggers a migration. However, in the embodiment of the present invention, it is considered that due to the existence of various types of bursty loads, such a method for specifying a threshold necessarily causes frequent virtual machine migration; therefore, in the embodiment of the present invention, when it is determined that the first load of the first physical machine at the first time is greater than the first threshold, the virtual machine migration is not directly performed, but the loads of a past period of time and a current time are comprehensively considered, so as to predict the future load, and the virtual machine migration is triggered only when the future load meets the preset condition, so that the interference of the bursty load on the system can be effectively prevented, and the performance loss of the system caused by unnecessary migration is reduced.

In practical application, the system can judge whether to trigger one virtual machine migration or not by combining the prediction result and the current load information. In practical application, firstly, whether the current load of a certain physical machine exceeds a threshold value k1 is analyzed, and if the current load exceeds the threshold value k1, the next step is executed; secondly, in order to verify whether the current load information belongs to the sudden load, the load at the future moment can be predicted based on the load data of the physical machine; then, after obtaining the load at the future time, if the load at the future time is greater than the threshold k1, the system will trigger a migration. Otherwise, the system does not trigger the migration, so that the interference of the sudden load to the system can be effectively avoided.

It should be noted that, by using the method for determining a virtual machine migration opportunity in the embodiment of the present invention, as long as a virtual machine migration command is triggered, some virtual machines on the physical machine must be selected for migration, the virtual machines that need to be migrated are referred to as candidate virtual machines, that is, virtual machines to be migrated, and the selection of the virtual machines to be migrated must meet a series of requirements, such as: the migration is short, and migration shock cannot occur to the system.

In the embodiment of the present invention, when the system selects the virtual machine to be migrated through the virtual machine to be migrated selection module, the system may be implemented through the following two steps:

step B1: and analyzing the resource type with the maximum utilization rate at present according to the predicted result.

Step B2: and D, selecting a virtual machine with the maximum utilization rate for the resource type analyzed in the step B1, wherein the virtual machine is the virtual machine to be migrated.

In another embodiment of the present invention, when the physical machine is overloaded, that is, the CPU utilization of the physical machine exceeds the first parameter, and when the physical machine is underloaded, that is, the CPU utilization of the physical machine is lower than the second parameter, it is necessary to migrate part or all of the virtual machines on the host to other hosts for operation according to the current state; the first parameter is greater than the second parameter, for example, the first parameter is 80% and the second parameter is 20%.

In the embodiment of the invention, when the virtual machine is in an overload state, the criterion for selecting the virtual machine to be migrated can be that the virtual probability with the maximum predicted load on the physical machine is migrated out of the current working host machine, namely the current physical machine, and the current physical machine can meet the future resource demand of the remaining virtual machine. When the virtual machine is in the low-load state, all the virtual machines are migrated out of the current physical machine, and then the virtual machine is deleted, so that the computing resources are saved.

And 103, determining a second physical machine, and migrating the virtual machine to be migrated to the second physical machine.

The second physical machine is different from the first physical machine, and the second physical machine belongs to the cloud computing management platform.

In the embodiment of the present invention, when the virtual machine to be migrated has been determined, the migration destination host, i.e., the destination physical machine, is selected next. The selection of the migration destination host needs to consider that sufficient resources can be provided for the virtual machine to be migrated, and it is ensured that after the migration is completed, the load of the destination host does not exceed the preset threshold value, which results in two or more migrations.

In the embodiment of the invention, when the system selects the migration target host through the target physical machine selection module, the system can be realized through the following three steps:

step C1: the destination host with the smaller load information is selected.

Step C2: when the number of the destination hosts is plural, as long as the destination host having the minimum load information is selected, the load of the destination host at a future time is predicted using the workload prediction method. If the load of the destination host at the future time is still minimal compared to the other destination hosts, the next step is performed, otherwise return to step C1 to reselect the destination host.

Step C3: if the load value of the virtual machine to be migrated plus the predicted value in step C2 does not exceed the preset threshold, the destination host is selected as the migration target of the virtual machine, otherwise, the process returns to step C1.

According to the virtual machine migration method provided by the embodiment of the invention, if the first load of the first physical machine is larger than the first threshold value at the current first moment, the second load of the first physical machine is predicted at the second moment; wherein the second time is after the first time; if the second load is larger than a second threshold value, selecting a virtual machine to be migrated from the virtual machines on the first physical machine; wherein the second threshold is greater than the first threshold; determining a second physical machine, and migrating the virtual machine to be migrated to the second physical machine; wherein the second physical machine is different from the first physical machine; that is to say, in the embodiment of the present invention, when it is determined that the bursty load of the first physical machine at the first time is greater than the first threshold, the load of the first physical machine at a future time is predicted, and when it is determined that the load at the future time is higher than the second threshold that is greater than the first threshold, a migration is triggered, so that interference of the bursty load on the system is effectively avoided, the problem that the virtual machine migration is triggered when the instantaneous bursty load reaches the fixed threshold in the related art, so that frequent virtual machine migration is caused, and unnecessary migration overhead is caused is solved, interference of the bursty load on the system is avoided, and reliability of the system is improved.

Based on the foregoing embodiments, the embodiment of the present invention provides a virtual machine migration method, which is applied to a cloud computing management platform; virtual machine clusters are established on physical machines of a cloud computing management platform, and the number of virtual machines to be migrated is at least two. In the embodiment of the invention, the virtual machine migration process can be optimized by means of a cluster deployment technology.

In the embodiment of the invention, the main function to be completed by the virtual machine cluster deployment method based on resource classification is to select N physical nodes, and the resources of the nodes are enough to deploy clusters of N virtual machines, so that the resources of the whole physical machine system are relatively balanced and comprehensively utilized. Referring to fig. 7 and 8, the method includes:

step 201, if the first load of the first physical machine at the current first time is greater than the first threshold, predicting a second load of the first physical machine at a second time.

Wherein the second time is after the first time.

In step 201, before predicting the second load of the first physical machine at the second time if the first load of the first physical machine at the first time is greater than the first threshold, the embodiment of the present invention further includes the following steps:

step S1: the cloud computing management platform analyzes the resource types of the virtual machine cluster, distinguishes the resource types, and has multiple modes for analyzing the cluster resource types: can be according to the user's specification; analysis may be based on historical data of the image used by the cluster.

Step S2: the cloud computing management platform filters the cluster deployment request, and if the cluster deployment request cannot be completed due to the fact that the limitation of a CPU, an internal memory, file capacity, network bandwidth and the like of a deployment system is analyzed, the deployment request should be filtered and fed back to a user.

Step 202, if the second load is greater than the second threshold, selecting a virtual machine to be migrated from the virtual machines on the first physical machine.

Wherein the second threshold is greater than the first threshold.

In the embodiment of the present invention, in the process of selecting the virtual machines to be migrated, it may be assumed that the number of the virtual machines is N, the number of the virtual machines to be deployed for initializing each physical machine node at this time is 0, and a priority queue with a capacity of N is created, where the priority queue represents a virtual machine resource load priority queue.

Step 203, determine the second physical machine.

In the embodiment of the invention, in the process of determining the second physical machine, all the physical machine nodes can be traversed circularly, whether a new virtual machine can be deployed at the physical machine node is calculated according to the following formula, and when the load information value L is less than L1, the current load ratio of the physical node is indicated to be low and can be used for deploying the virtual machine; when the load information value L1< L < L2 indicates that the current load of the node of the physical machine is in the best state, the virtual machine can be deployed, but the newly added virtual machine cannot cause the load of the node of the physical machine to be too heavy; when the load information value L > L2, it indicates that the physical machine node is currently overloaded, and the physical machine node should be reselected.

In the embodiment of the invention, the load information is determined according to the resource type of the cluster, and the cluster is divided into a calculation intensive cluster, a storage intensive cluster and a flow intensive cluster. Furthermore, the virtual machine to be migrated is determined based on the cluster, so that the positioning efficiency can be improved.

In the embodiment of the invention, the load value is calculated according to the resource type of the cluster and the weight of different resources, and then the load priority queue of the virtual machine resource is inserted until physical nodes are selected for all the virtual machines.

And step 204, acquiring the load of each virtual machine to be migrated.

Step 205, determining a target physical machine corresponding to each virtual machine to be migrated from the plurality of second physical machines.

Wherein the plurality of second physical machines includes a target physical machine.

And step 206, acquiring the loads of the plurality of target physical machines.

Step 207, sequencing the plurality of target physical machines according to the load of each virtual machine to be migrated and the load of each target physical machine to obtain a priority queue.

In this embodiment of the present invention, step 207 ranks, according to the load of each virtual machine to be migrated and the load of each target physical machine, the plurality of target physical machines to obtain the priority queue, and may be implemented by the following steps:

step 207a, calculating the sum of the load of each virtual machine to be migrated and the load of the target physical machine corresponding to each virtual machine to be migrated, and obtaining the transmission load.

And step 207b, sequencing the target physical machines according to the transmission load to obtain a priority queue.

And step 208, determining the priority of each target physical machine according to the priority queue.

And 209, carrying out mirror image transmission on the virtual machines to be migrated corresponding to each target physical machine according to the sequence of the priorities from high to low.

And the transmission load corresponding to the target physical machine with the high priority is smaller than the transmission load corresponding to the target physical machine with the low priority.

In the embodiment of the invention, the virtual machine to be migrated is migrated from the first physical machine to the target physical machine in the second physical machine, namely the virtual machine is recovered, the process consumes the resources of the physical machine, and if the current load of the target physical machine is higher, the virtual machine is waited to be retransmitted; that is, the target physical machine has a low load, and the virtual machine is restored based on the mirror image after the mirror image is transferred. And finally, starting all the virtual machines of the cluster to complete cluster deployment.

Based on the foregoing embodiments, the embodiments of the present invention provide a virtual machine migration method, which is applied to a cloud computing management platform, and it should be noted that this embodiment aims to adopt a virtual machine cluster mirror image fast express technology in a virtual machine migration process to assist in improving the efficiency of virtual machine dynamic migration.

Because the virtual machine image file contains cluster application, an operating system and the like, the file size of the virtual machine image file is often tens of millions of bytes and even hundreds of millions of bytes. Therefore, it may take a long time to transfer an image file from an image library to a physical machine, while a virtual machine cluster may consist of tens of virtual machines, and a greater time penalty is required to deploy and migrate such a cluster. The mirror image rapid transfer method takes the mirror image and the copy of a mirror image library as root nodes, after the first copy is completed, the number of the mirror image and the copy of the mirror image file which is just transferred in the whole cloud platform system is twice as large as the number of the mirror image and the copy, then the mirror image and the copy in the mirror image library and the mirror image which is transferred for the first time are taken as father nodes of the second mirror image transfer, and after the second mirror image transfer is completed, the number of the mirror image and the copy of the mirror image file in the system is four times as large as the number of the. And the transmission is circulated until the number of the virtual machines required by the cluster can be met. For example, before the first copy in fig. 8, the system only includes image files such as image1 and image3, image2 and image6, and image4 and image5 in solid frames; after the first copy is completed, the system comprises image files such as image2 and image4, image1 and image5, and image3 and image6 in a dotted line frame, namely, after the first copy is completed, the image files just transferred in the whole cloud platform system share twice the number of images and copies.

In the embodiment of the present invention, the virtual machine cluster mirror image fast express technology may include the following steps:

firstly: assuming that M backup exists in a mirror image in the mirror image library, the mirror image transfer strategy picks out M physical machines from the selected physical machine, and then transfers the mirror image file to the selected physical machine, which is basically a mirror image backup to one physical machine.

Secondly, the method comprises the following steps: when this transfer is complete, the M copies in the mirror library are referred to as root nodes, and the M physical machines that have received the mirror are referred to as first generation child nodes.

And thirdly: and if the deployment request of the virtual machine node is not completed, respectively sending the mirror image to 2M second-generation child nodes by using M backup nodes, namely the root node and M first-generation child nodes.

And finally: and circulating the above steps until the deployment request is completed.

Therefore, the virtual machine cluster image rapid express technology can provide policy support for multi-path parallel transmission of images, and transmission speed of virtual machine cluster image files is greatly improved.

It should be noted that, for the descriptions of the same steps and the same contents in this embodiment as those in other embodiments, reference may be made to the descriptions in other embodiments, which are not described herein again.

Based on the foregoing embodiments, an embodiment of the present invention provides a virtual machine migration method, which is applied to a cloud computing management platform, and the method includes the following steps:

step 301, receiving a user request, and obtaining a request type of the user request.

Wherein the request type comprises a static request or a dynamic request.

In practical applications, each server cluster system has its own load characteristics, and the load indexes to be considered are also different, for example: in a Web server cluster system, for a request of static content, a server only needs to read corresponding content from a disk and send the corresponding content to a browser; when a request is made for dynamic content, the request file is often generated and sent to the client after reading, compiling and processing.

Referring to fig. 9, in the embodiment of the present invention, the user requests may be divided into two types, where, for the requests of static content, the main load of such requests on the server is the network bandwidth load. On Web applications, such requests are primarily requests for static Web pages. For dynamic content requests, the load brought to the server by such requests is mainly the occupation of CPU and memory resources. In Web applications, such requests are primarily requests for dynamic Web pages and access to databases.

It should be noted that, when determining the request type requested by the user, the determination may be performed according to the file type requested by the user.

Step 302, if the first load is greater than the first threshold, a third load at least two moments in a preset time period is obtained.

The preset time period comprises a first moment and at least one third moment before the first moment, and the third load at each moment of at least two moments comprises at least two loads with different attributes; the loads of different attributes characterize different types of resources occupied in the first physical machine.

In an embodiment of the present invention, the loads with at least two different attributes include: at least two of the utilization rate of the central processing unit of the first physical machine, the utilization rate of the memory of the first physical machine, the utilization rate of the input interface/output interface of the first physical machine, and the utilization rate of the network bandwidth of the first physical machine.

And step 303, acquiring the attribute of each load in the third loads at each moment.

In the embodiment of the invention, the loads with different attributes represent different types of occupied resources in the first physical machine.

In this embodiment of the present invention, the third load at each time may include at least two of a utilization rate of a central processing unit of the first physical machine, a utilization rate of a memory of the first physical machine, a utilization rate of an input interface/output interface of the first physical machine, and a utilization rate of a network bandwidth of the first physical machine.

Step 304, determining a weight value of each load in the third loads at each moment according to the request type and the attribute of each load in the third loads at each moment.

In the embodiment of the invention, the predicted value of the working load at the next moment is calculated by weighted average according to the current moment and the previous load data, when the weight of the load data is determined, the type of a user request needs to be judged firstly, if the request is a request for static content, the weight of a network bandwidth item is increased by 20%, and meanwhile, the weights of a CPU and a memory are both reduced by 10%; if the request is for dynamic content, the weight of both the CPU and the memory is increased by 10%, and the weight of the network bandwidth item is reduced by 20%.

For example, when the user request is a static request, according to the request type and the attribute of each load in the third load at a certain time, a weight value of each load in the third load is determined, for example, the weight value of the CPU utilization rate is 15%, the weight value of the memory utilization rate is 15%, the weight value of the input interface/output interface utilization rate is 25%, and the weight value of the network utilization rate bandwidth is 45%.

When the user request is a dynamic request, determining a weight value of each load in the third load according to the request type and the attribute of each load in the third load at a certain moment, for example, the weight value of the CPU utilization rate is 35%, the weight value of the memory utilization rate is 35%, the weight value of the input interface/output interface utilization rate is 25%, and the weight value of the network utilization rate and the bandwidth is 5%.

Step 305, calculating a fourth load at each moment according to the third load at each moment and the weight value of the third load at each moment.

In the embodiment of the present invention, the fourth load at each time may be calculated by using formula 1.

L-a 1 × ucpu + a2 × umem + a3 × uio + a4 × unet (formula 1)

Wherein L represents a fourth load at each time; ucpu represents the CPU utilization rate, umem represents the memory utilization rate, uio represents the I/O utilization rate, and unet represents the network bandwidth utilization rate; a1 represents a CPU utilization rate weight value, a2 represents a memory utilization rate weight value, a3 represents an I/O utilization rate weight value, a4 represents a network bandwidth utilization rate weight value, and a1+ a2+ a3+ a4 is equal to 1.

And step 306, calculating a second load according to the fourth load at each moment.

In the embodiment of the invention, the average value of the fourth loads at a plurality of moments is calculated to obtain the second load.

And 307, if the second load is greater than the second threshold, selecting a virtual machine to be migrated from the virtual machines on the first physical machine.

Wherein the second threshold is greater than the first threshold.

In this embodiment of the present invention, if the second load is greater than the second threshold in step 307, selecting a virtual machine to be migrated from virtual machines on the first physical machine, may include the following steps:

step 307a1, if the second load is greater than the second threshold, obtaining the resource utilization rate of the first physical machine.

Step 307b1, selecting a virtual machine to be migrated from the plurality of virtual machines on the first physical machine according to the resource utilization rate.

In this embodiment of the present invention, step 307b1 is to select a virtual machine to be migrated from multiple virtual machines on the first physical machine according to the resource utilization rate, and may further include the following steps:

and step one, if the resource utilization rate is larger than or equal to a third threshold value, predicting a fifth load of each virtual machine on the first physical machine at a third moment.

Wherein the third time is after the first time.

In the embodiment of the present invention, if the resource utilization rate is greater than or equal to the third threshold, it is determined that the physical machine is overloaded, and the fifth load of each virtual machine on the first physical machine at the third time is predicted.

And secondly, selecting a virtual machine corresponding to the maximum load in the fifth load from the virtual machines on the first physical machine, and determining the virtual machine as a virtual machine to be migrated.

In this embodiment of the present invention, if the second load is greater than the second threshold in step 307, selecting a virtual machine to be migrated from virtual machines on the first physical machine, which may further include the following steps:

step 307a2, if the resource utilization rate is less than or equal to the fourth threshold, determining each virtual machine on the first physical machine as the virtual machine to be migrated.

Wherein the fourth threshold is less than the third threshold.

In the embodiment of the present invention, if the resource utilization rate is less than or equal to the fourth threshold, it is determined that the physical machine has a low load condition, and it is further determined that all virtual machines on the physical machine are to-be-migrated virtual machines.

In practical applications, when an overload condition occurs to a physical machine, such as the CPU utilization exceeds 80%, and when the CPU utilization is lower than 20%, some or all of the virtual machines on the physical machine need to be migrated to other physical machines to run according to the current state. And when the virtual machines are in an overload state, the virtual machines are arranged in a descending order according to the future workload, and the virtual machines with the top future workload rank are selected for migration. The selection criterion is to migrate the virtual probability of the highest predicted load out of the current physical machine first and enable the host to meet the future resource requirements of the remaining virtual machines. When in the underloaded state, all virtual machines are migrated out of the current physical machine.

And 308, determining the second physical machine, and migrating the virtual machine to be migrated to the second physical machine.

Wherein the second physical machine is different from the first physical machine.

In this embodiment of the present invention, the determining of the second physical machine in step 308 may be implemented by the following steps:

step 308a, acquiring a sixth load of each physical machine in the cloud computing management platform at the first moment.

And 308b, selecting a physical machine with the sixth load smaller than a fifth threshold value from the physical machines in the cloud computing management platform, and determining the physical machine as a third physical machine.

The third physical machine is different from the first physical machine, and the number of the third physical machines is multiple.

In the embodiment of the invention, when the system determines the third physical machine, the physical machine with the minimum load can be selected from the physical machines in the cloud computing management platform and determined as the first and third physical machines. And further executing step 308c based on the first third physical machine, if the predicted load, namely the seventh load, corresponding to the first third physical machine is not the minimum, then reselecting the minimum physical machine from the physical machines in the cloud computing management platform as the second third physical machine, and executing the subsequent steps.

And step 308c, predicting a seventh load of the virtual machine on each third physical machine at the fourth moment.

Wherein the fourth time is after the first time.

And 308d, calculating the sum of the seventh load of each third physical machine and the load of the virtual machine to be migrated to obtain a plurality of target loads.

And step 308e, determining the target load smaller than the sixth threshold value in the plurality of target loads.

And 308f, selecting a third physical machine corresponding to the target load smaller than the sixth threshold from the third physical machines, and determining the third physical machine as a fourth physical machine.

Wherein the number of the fourth physical machines is at least one.

And 308g, determining a second physical machine from at least one fourth physical machine.

In this embodiment of the present invention, the step 308g of determining the second physical machine from the at least one fourth physical machine may be implemented by the following steps:

firstly, calculating a value obtained by subtracting the sixth load of each fourth physical machine from the seventh load of each fourth physical machine to obtain the resource demand scaling quantity of each fourth physical machine.

In the embodiment of the invention, the resource demand scaling quantity is a negative value, which indicates that the future load is reduced. The resource demand scale is positive, indicating that future loads are increasing.

And secondly, calculating the total resource amount of each fourth physical machine minus the value of the sixth load of each fourth physical machine to obtain the unallocated resource amount of each fourth physical machine.

And thirdly, calculating the value obtained by subtracting the resource demand scaling quantity of each fourth physical machine from the unallocated resource quantity of each fourth physical machine to obtain the predicted residual resource quantity of each fourth physical machine.

And finally, determining the second physical machine from the fourth physical machines according to the predicted residual resource amount of each fourth physical machine.

In the embodiment of the present invention, determining the second physical machine from the fourth physical machines according to the predicted remaining resource amount of each fourth physical machine may be implemented by the following steps: and determining the fourth physical machine corresponding to the maximum value in the predicted residual resource amount of the fourth physical machine as the second physical machine.

step 401, if the first load is greater than the first threshold, obtaining third loads of at least two moments in a preset time period.

The preset time period comprises a first moment and at least one third moment before the first moment, and the third load of each moment in the at least two moments comprises at least two loads with different attributes.

Referring to fig. 10, in the embodiment of the present invention, the third load at each time includes a distance, where the distance represents a distance between the first physical machine and the data center that receives the user request, and the cloud computing management platform further includes a data center, where the data center is configured to manage a third physical machine, and the third physical machine includes the first physical machine; meanwhile, the third load at each moment further includes at least two of the utilization rate of the central processing unit of the first physical machine, the utilization rate of the memory of the first physical machine, the utilization rate of the input interface/output interface of the first physical machine, and the utilization rate of the network bandwidth of the first physical machine.

It should be noted that a large cloud computing platform usually includes a plurality of data centers, so-called city-sharing data centers, the plurality of data centers are not generally located in one building, but are distributed in data center rooms at different locations, the distance between each location is 5-10 kilometers, if a virtual machine migrates across the data centers, due to the influence of the distance, service level agreements of businesses are affected, and thus migration of the virtual machine across the data centers is limited.

In the embodiment of the invention, the predicted value of the working load at the next moment is calculated by weighted average according to the load data at the current first moment and a plurality of second moments before the current first moment, and when the weighted average is calculated, distance indexes are added except four indexes of a CPU (Central processing Unit), a memory, an I/O (input/output) and a network bandwidth, and the unit can be kilometers. The effect of this implementation is that the predicted load of the local data center is low, the load of the remote data center is high, and the virtual machine is effectively prevented from migrating to the remote data center.

And step 402, acquiring the attribute of each load in the third loads at each moment.

Step 403, determining a weight value of each load in the third loads at each time according to the attribute of each load in the third loads at each time.

Referring to fig. 10, for example, the CPU utilization rate is 20%, the memory utilization rate is 20%, the input interface/output interface utilization rate is 20%, and the network utilization bandwidth is 20%.

And step 404, calculating a fourth load at each moment according to the third load at each moment and the weight value of the third load at each moment.

Step 405, calculating a second load according to the fourth load at each moment.

And 406, if the second load is greater than the second threshold, selecting a virtual machine to be migrated from the virtual machines on the first physical machine.

Wherein the second threshold is greater than the first threshold. In practical applications, when an overload condition occurs to a physical machine, such as the CPU utilization exceeds 80%, and when the CPU utilization is lower than 20%, some or all of the virtual machines on the physical machine need to be migrated to other physical machines to run according to the current state. And when the virtual machines are in an overload state, the virtual machines are arranged in a descending order according to the future workload, and the virtual machines with the top future workload rank are selected for migration. The selection criterion is to migrate the virtual probability of the highest predicted load out of the current physical machine first and enable the host to meet the future resource requirements of the remaining virtual machines. When in the underloaded state, all virtual machines are migrated out of the current physical machine.

Step 407, determining the second physical machine, and migrating the virtual machine to be migrated to the second physical machine.

In summary, the embodiment of the invention has the following beneficial effects:

(1) the method for predicting the working load is provided, the load comprehensively considers a plurality of factors such as CPU occupancy rate, memory occupancy rate, disk IO, network bandwidth and distance, the future load capacity of the task executed by the virtual machine is predicted, and the weight of a load index is adjusted by classifying the service request, so that the virtual machine to be migrated and the host computer are accurately screened out, the migration frequency of the virtual machine is reduced, and migration oscillation is avoided.

(2) The virtual machine cluster deployment method based on resource classification is provided, different virtual machine clusters are divided according to resource types, so that the virtual machines to be migrated and the target physical host can be rapidly screened out, and the dynamic migration efficiency of the virtual machines is assisted to be improved.

(3) The method takes the mirror image and the copy of a mirror image library as root nodes, and through repeated cyclic copying, the mirror image of the virtual machine can be quickly copied at the speeds of M, 2M and 4M.

(4) The method for preventing the virtual machine from migrating across the data centers is provided, and migration of the virtual machine under a multi-data center scene is effectively prevented by introducing the data center distance index into calculation of load prediction, so that a service level agreement of a service is ensured.

Based on the foregoing embodiments, an embodiment of the present invention provides a cloud computing management platform, where the cloud computing management platform may be applied to the virtual machine migration method provided in the embodiments corresponding to fig. 5 and 7, and referring to fig. 11, the cloud computing management platform 11 includes: a processor 1101, a memory 1102, and a communication bus 1103, wherein:

the communication bus 1103 is used to enable communication connections between the processor 1101 and the memory 1102.

Processor 1101 is configured to execute a virtual machine migration program stored in memory 1102 to implement the following steps:

if the first load of the first physical machine at the current first moment is larger than a first threshold value, predicting a second load of the first physical machine at a second moment; wherein the second time is after the first time;

In other embodiments of the present invention, the processor 1101 is configured to execute the following steps when predicting the second load of the first physical machine at the second time if the first load of the first physical machine at the current first time is greater than the first threshold in the memory 1102:

if the first load is larger than the first threshold, acquiring a third load of at least two moments in a preset time period; the preset time period comprises a first moment and at least one third moment before the first moment, and the third load at each moment of at least two moments comprises at least two loads with different attributes; the loads with different attributes represent different types of occupied resources in the first physical machine;

and calculating the second load according to the fourth load at each moment.

In other embodiments of the present invention, when the processor 1101 is configured to execute the second load calculated according to the fourth load at each moment in time in the memory 1102, the following steps may be implemented:

and calculating the average value of the fourth loads at a plurality of moments to obtain a second load.

In other embodiments of the present invention, processor 1101 is configured to execute a virtual machine migration program in memory 1102 to implement the following steps:

correspondingly, before the fourth load at each moment is calculated according to the third load at each moment and the weight value of the third load at each moment, the method further includes:

acquiring the attribute of each load in the third loads at each moment;

and determining the weight value of each load in the third loads at each moment according to the request type and the attribute of each load in the third loads at each moment.

In other embodiments of the present invention, the third load at each time further includes a distance, where the distance represents a distance between the first physical machine and the data center that receives the user request, and the cloud computing management platform further includes a data center, where the data center is configured to manage a third physical machine, and the third physical machine includes the first physical machine.

acquiring the attribute of each load in the third loads at each moment;

and determining the weight value of each load in the third loads at each moment according to the attribute of each load in the third loads at each moment.

In other embodiments of the present invention, the processor 1101 is configured to execute the memory 1102, where the number of virtual machines on the first physical machine is multiple, and if the second load is greater than the second threshold, when selecting a virtual machine to be migrated from the virtual machines on the first physical machine, the following steps may be implemented:

if the second load is larger than a second threshold value, acquiring the resource utilization rate of the first physical machine;

In other embodiments of the present invention, when the processor 1101 is configured to execute the selection of the virtual machine to be migrated from the plurality of virtual machines on the first physical machine in the memory 1102 according to the resource utilization rate, the following steps may be implemented:

if the resource utilization rate is greater than or equal to a third threshold value, predicting a fifth load of each virtual machine on the first physical machine at a third moment; wherein the third time is after the first time;

and selecting the virtual machine corresponding to the maximum load in the fifth load from the virtual machines on the first physical machine, and determining the virtual machine as the virtual machine to be migrated.

if the resource utilization rate is less than or equal to a fourth threshold value, determining each virtual machine on the first physical machine as a virtual machine to be migrated; wherein the fourth threshold is less than the third threshold.

In other embodiments of the present invention, when the processor 1101 is configured to execute the second physical machine determined in the memory 1102, the following steps may be implemented:

selecting a physical machine with a sixth load smaller than a fifth threshold value from the physical machines in the cloud computing management platform, and determining the physical machine as a third physical machine; the third physical machines are different from the first physical machines, and the number of the third physical machines is multiple;

predicting a seventh load of the virtual machine on each third physical machine at the fourth time; wherein the fourth time is after the first time;

a second physical machine is determined from the at least one fourth physical machine.

In other embodiments of the present invention, when the processor 1101 is configured to execute the step of determining the second physical machine from the at least one fourth physical machine in the memory 1102, the step of:

calculating the value of subtracting the resource demand expansion amount of each fourth physical machine from the unallocated resource amount of each fourth physical machine to obtain the predicted residual resource amount of each fourth physical machine;

In other embodiments of the present invention, when the processor 1101 is configured to execute the step of determining the second physical machine from the fourth physical machines according to the predicted remaining resource amount of the fourth physical machine in the memory 1102, the step of:

In other embodiments of the present invention, a virtual machine cluster is established on physical machines of a cloud computing management platform, the number of virtual machines to be migrated is at least two, the number of virtual machines to be migrated is the same as the number of second physical machines, and a processor 1101 is configured to execute a virtual machine migration program in a memory 1102, so as to implement the following steps:

acquiring the load of each virtual machine to be migrated;

determining a target physical machine corresponding to each virtual machine to be migrated from the plurality of second physical machines; wherein the plurality of second physical machines comprise target physical machines;

acquiring loads of a plurality of target physical machines;

In other embodiments of the present invention, when the processor 1101 is configured to execute the sorting of the plurality of second physical machines according to the load of each virtual machine to be migrated and the load of each target physical machine in the memory 1102 to obtain the priority queue, the method may be implemented by:

In other embodiments of the present invention, when the processor 1101 is configured to execute the root in the memory 1102 to migrate the virtual machine to be migrated to the second physical machine, the following steps may be implemented:

It should be noted that, for a specific implementation process of the step executed by the processor in this embodiment, reference may be made to the implementation process in the virtual machine migration method provided in the embodiments corresponding to fig. 5 and 7, and details are not described here again.

Based on the foregoing embodiments, the embodiments of the present invention provide a storage medium, where one or more programs are stored, and the one or more programs are executable by one or more processors to implement the steps in the virtual machine migration method provided in the embodiments corresponding to fig. 5 and 7.

The computer storage medium may be a Memory such as a Read Only Memory (ROM), a Programmable Read Only Memory (PROM), an Erasable Programmable Read Only Memory (EPROM), an Electrically Erasable Programmable Read Only Memory (EEPROM), a magnetic Random Access Memory (FRAM), a Flash Memory (Flash Memory), a magnetic surface Memory, an optical Disc, or a Compact Disc Read-Only Memory (CD-ROM); and may be various electronic devices such as mobile phones, computers, tablet devices, personal digital assistants, etc., including one or any combination of the above-mentioned memories.

It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, recitation of an element by the phrase "comprising an … …" does not exclude the presence of other like elements in the process, method, article, or apparatus that comprises the element.

The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.

Through the above description of the embodiments, those skilled in the art will clearly understand that the method of the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but in many cases, the former is a better implementation manner. Based on such understanding, the technical solutions of the present invention may be embodied in the form of a software product, which is stored in a storage medium (such as ROM/RAM, magnetic disk, optical disk) and includes instructions for enabling a terminal device (such as a mobile phone, a computer, a server, an air conditioner, or a network device) to execute the method described in the embodiments of the present invention.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

The above description is only a preferred embodiment of the present invention, and not intended to limit the scope of the present invention, and all modifications of equivalent structures and equivalent processes, which are made by using the contents of the present specification and the accompanying drawings, or directly or indirectly applied to other related technical fields, are included in the scope of the present invention.

Claims

1. A virtual machine migration method is applied to a cloud computing management platform, a virtual machine cluster is established on a physical machine of the cloud computing management platform, and the method comprises the following steps:

if the second load is larger than a second threshold value, selecting a virtual machine to be migrated from the virtual machines on the first physical machine; the second threshold value is larger than the first threshold value, and the number of the virtual machines to be migrated is at least two;

determining a second physical machine, wherein the second physical machine is different from the first physical machine, and the number of the virtual machines to be migrated is the same as that of the second physical machine;

acquiring the load of each virtual machine to be migrated;

acquiring loads of a plurality of target physical machines;

sequencing the target physical machines according to the transmission load to obtain a priority queue;

2. The method of claim 1, wherein predicting a second load of the first physical machine at a second time if a first load of the first physical machine at a current first time is greater than a first threshold comprises:

and calculating the second load according to the fourth load at each moment.

3. The method of claim 2, wherein said calculating the second load according to the fourth load at each time comprises:

4. The method of claim 2, wherein before obtaining the plurality of different types of load data for each of the plurality of time instants prior to the first time instant if the first load is greater than the first threshold, the method further comprises:

acquiring the attribute of each load in the third loads at each moment;

5. The method of claim 2, wherein the third load at each time further comprises a distance, wherein the distance characterizes a distance between the first physical machine and a data center that receives the user request, wherein the cloud computing management platform further comprises the data center, wherein the data center is configured to manage a third physical machine, and wherein the third physical machine comprises the first physical machine.

6. The method according to claim 5, wherein before the calculating the fourth load at each time according to the third load at each time and the weight value of the third load at each time, the method further comprises:

acquiring the attribute of each load in the third loads at each moment;

7. The method of claim 1, wherein the number of virtual machines on the first physical machine is multiple, and the selecting the virtual machine to be migrated from the virtual machines on the first physical machine if the second load is greater than a second threshold comprises:

8. The method of claim 7, wherein selecting a virtual machine to be migrated from the plurality of virtual machines on the first physical machine based on the resource utilization comprises:

9. The method of claim 7, wherein selecting a virtual machine to be migrated from the plurality of virtual machines on the first physical machine based on the resource utilization comprises:

10. The method of claim 1, wherein determining the second physical machine comprises:

11. The method of claim 10, wherein the determining the second physical machine from among the at least one fourth physical machine comprises:

12. The method of claim 11, wherein determining the second physical machine from the fourth physical machines based on the predicted remaining amount of resources of the fourth physical machines comprises:

13. A cloud computing management platform, the cloud computing management platform comprising:

a memory for storing executable instructions;

a processor for executing executable instructions stored in the memory to implement the virtual machine migration method of any of claims 1 to 12.

14. A storage medium storing one or more programs, the one or more programs being executable by one or more processors to implement the steps of the virtual machine migration method according to any one of claims 1 to 12.