Resource management and control method and device and storage medium

By monitoring and managing the resource status on the physical host, the problem of resource idleness is solved, resource utilization is improved, and the timely deployment of instances is ensured.

CN119728691APending Publication Date: 2025-03-28HANGZHOU ALICLOUD FEITIAN INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202311289217.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2023-09-28
Publication Date
2025-03-28

AI Technical Summary

Technical Problem

In hybrid deployment solutions, resource idle issues on physical hosts lead to waste of resources, especially scarce resources such as GPUs, resulting in higher cost losses.

Method used

By monitoring the resource status of the specified type of resources on the target physical host and the target type resources they need to match, if the specification guarantee requirements are not met, the target instance is selected from the target physical host and moved out to ensure that the resource status meets the specification guarantee requirements.

Benefits of technology

It effectively improves resource utilization on the physical host, avoids the situation where the specified type of resources are idle due to insufficient target type resources, and ensures timely creation and deployment of instances.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119728691A_ABST
    Figure CN119728691A_ABST
Patent Text Reader

Abstract

The embodiment of the invention provides a resource management and control method and device and a storage medium. The resource states of specified type resources on a target physical host and target type resources needing to be matched with the specified type resources can be monitored respectively; if the monitored resource state does not meet the specification guarantee requirement set for the specified type of resource, selecting a target instance from instances occupying the target type of resource on the target physical host; and migrating the target instance from the target physical host. Thus, the target type resources on the target physical host can be supplemented in time through resource management and control, the situation that the specified type resources have to be idle due to the fact that the remaining amount of the specified type resources is sufficient and the remaining amount of the target type resources needing to be matched is insufficient is avoided, the utilization rate of the specified type resources is effectively increased, and the resource utilization rate is increased. And various instances needing to occupy the specified type of resources can be created more timely.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of cloud computing technology, and in particular to a resource management method, device and storage medium. Background Art

[0002] The hybrid deployment solution supports the coexistence of multiple different types of instances on the same physical host. Different types of instances may require different types of resources and resource specifications.

[0003] Currently, in hybrid deployment solutions, idle resources often occur, resulting in a waste of physical resources, especially scarce resources such as GPUs (Graphics Processing Units), which will result in high cost losses if left idle. Summary of the invention

[0004] Multiple aspects of the present application provide a resource management method, device and storage medium to improve resource utilization on a physical host.

[0005] The present application provides a resource management method, including:

[0006] Monitor the resource status of the specified type of resources and the target type of resources required to match them on the target physical host respectively;

[0007] If the monitored resource status does not meet the specification guarantee requirement set for the specified type of resource, selecting a target instance from instances occupying the target type of resource on the target physical host;

[0008] The target instance is migrated from the target physical host so that the resource states of the designated type of resources and the target type of resources in the target physical host meet the specification guarantee requirement.

[0009] The embodiment of the present application also provides a management and control device, including a memory, a processor, and a communication component;

[0010] The memory is used to store one or more computer instructions;

[0011] The processor is coupled to the memory and the communication component, and is configured to execute the one or more computer instructions to execute the aforementioned resource management and control method.

[0012] The embodiment of the present application further provides a physical host, including a memory, a processor, and a communication component, wherein the physical host provides resources of a specified type and resources of a target type and carries one or more instances;

[0013] The memory is used to store one or more computer instructions;

[0014] The processor is coupled to the memory and the communication component and is configured to execute the one or more computer instructions for:

[0015] Monitoring the resource status of the designated type of resources and the target type of resources required to be matched;

[0016] If the monitored resource status does not meet the specification guarantee requirement set for the specified type of resource, a target instance is selected from instances occupying the target type of resource;

[0017] The target instance is migrated out so that the resource states of the designated type resources and the target type resources meet the specification guarantee requirement.

[0018] An embodiment of the present application also provides a computer-readable storage medium storing computer instructions. When the computer instructions are executed by one or more processors, the one or more processors are caused to execute the aforementioned resource management method.

[0019] In an embodiment of the present application, the resource status of the specified type of resource and the target type of resource required to be matched on the target physical host can be monitored separately; if the monitored resource status does not meet the specification guarantee requirements set for the specified type of resource, the target instance is selected from the instances occupying the target type of resource on the target physical host; the target instance is migrated from the target physical host so that the resource status of the specified type of resource and the target type of resource in the target physical host meet the specification guarantee requirements. In this way, the target type of resource on the target physical host can be replenished in time through resource management and control, avoiding the situation where the remaining amount of the specified type of resource is sufficient, while the remaining amount of the target type of resource required to be matched is insufficient, resulting in the specified type of resource having to be idle, thereby effectively improving the utilization rate of the specified type of resource and ensuring that various instances that need to occupy the specified type of resource can be created more timely. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0021] Figure 1 A logical schematic diagram of a resource management method provided for an exemplary embodiment of the present application;

[0022] Figure 2 A schematic diagram of a resource management method provided as a flowchart of an exemplary embodiment of the present application;

[0023] Figure 3 A flowchart of a resource management method provided for another exemplary embodiment of the present application;

[0024] Figure 4 A schematic diagram of a process flow of a resource management and control method provided for another exemplary embodiment of the present application;

[0025] Figure 5a and 5b A schematic diagram of an exemplary application scenario provided for this application;

[0026] Figure 6 A schematic diagram of the structure of a control device provided for an exemplary embodiment of the present application;

[0027] Figure 7 A schematic diagram of the structure of a physical host provided as another exemplary embodiment of the present application. DETAILED DESCRIPTION

[0028] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.

[0029] Before describing in detail the technical solutions provided by the embodiments of the present application, several technical concepts involved in the present application are briefly described as follows.

[0030] Physical host: also known as physical server, is a physical machine hardware facility, including independent resources such as CPU, memory, disk, etc. Physical host, combined with the deployment of software environment, can provide various basic resources for instance, which is the instance running environment.

[0031] Cloud instance: It is the foundation of cloud computing and is used to provide computing services. It is a simulated server that can contain the most basic computing components such as CPU, memory, operating system, network, disk, etc. Cloud instances can include virtual machines, containers, etc.

[0032] Instance products: Instances are classified according to their performance, price, configuration and other characteristics to meet different usage scenarios of customers.

[0033] Instance specifications: The specifications of the various physical resources required by the instance, such as the number of CPU cores or memory capacity required.

[0034] Hybrid deployment: Under the same hardware and software environment, a single type of physical host can support the coexistence of multiple instance products. Hybrid deployment mainly involves two aspects of mixing. One aspect is that the physical resources provided on the physical host need to be multiple; the other aspect is that the physical host can support the deployment of various instance types, and the types of physical resources required by different instances may not be exactly the same. For example, a physical host can provide CPU resources and GPU resources at the same time, and can support the deployment of GPU instances and CPU instances, where the types of resources required by GPU instances and CPU instances are different.

[0035] During the research, the inventor discovered that creating or releasing an instance on a physical host will cause the resource status on the physical host to change. However, in many cases, the resource status on the physical host can no longer meet the specification requirements of the relevant instance. On the one hand, this will cause some of the remaining resources to become fragmented resources, which will be idle for a long period of time, resulting in a waste of resources; on the other hand, instances of relevant specifications may not be delivered on time due to the inability to find a qualified physical host, resulting in untimely delivery of instances. To this end, this embodiment proposes a resource management and control solution for managing fragmented resources on the physical host to improve resource utilization in the physical host.

[0036] Figure 1 A logical diagram of a resource management method provided by an exemplary embodiment of the present application. Figure 1 The resource management and control method provided in this embodiment can be executed by a management and control device. The management and control device can be implemented as software, hardware, or a combination of software and hardware. The management device can be integrated into the management and control equipment.

[0037] The control device in this embodiment can adopt the existing control node used for controlling the physical host in the cloud computing technology, and transform the existing control node so that it can execute the resource control method provided in this embodiment. In other words, a control module can be added to the existing control node to execute the resource control method provided in this embodiment. Of course, the control device in this embodiment can also be an independent device unrelated to the existing control node. In this case, the control device in this embodiment can communicate with each physical host that has been deployed to control the resources in each physical host.

[0038] In this embodiment, the types and amounts of physical resources included in different physical hosts may be different, and no limitation is made here.

[0039] The inventors found during the research that, usually, the resource types specified in the instance specifications are diverse. Therefore, various types of resources in the physical host usually need to be sold together to adapt to the instance specifications. To this end, this embodiment proposes to set specification guarantee requirements for the specified type of resources, and through resource management and control, the specified type of resources on the physical host can meet the specification guarantee requirements, thereby avoiding the aforementioned fragmentation problem of the specified type of resources.

[0040] The specification guarantee requirement in this embodiment is used to ensure that the specified type of resources and the target type of resources on the physical host can meet the specified specification requirements. The resource management method provided in this embodiment proposes to use resource management to enable the specified type of resources and the target type of resources on the physical host to meet the specification guarantee requirement, thereby managing the fragmentation problem of the specified type of resources on the physical host.

[0041] Figure 2 A schematic diagram of a resource management method provided by an exemplary embodiment of the present application. Figure 2 , the method may include:

[0042] Step 200: monitor the resource status of the specified type of resources and the target type of resources required to be matched on the target physical host respectively;

[0043] Step 201: If the monitored resource status does not meet the specification guarantee requirement set for the specified type of resource, a target instance is selected from the instances occupying the target type of resource on the target physical host;

[0044] Step 202: Migrate the target instance from the target physical host so that the resource states of the designated type resources and the target type resources in the target physical host meet the specification guarantee requirements.

[0045] In this embodiment, the timing of the resource control of the target physical host by the control device is not limited. For example, the control device can periodically control the resources of the target physical host. Of course, the control device can be triggered manually or at a specified time node to control the resources of the target physical host. In each round of resource control, the control device can implement the above steps 200 to 202 to manage the fragmented resources on the target physical host.

[0046] Among them, the specified type of resources in this embodiment can be any type of physical resource that needs to be fragmented. The specified type of resources in this embodiment can include but are not limited to GPU resources and local storage resources, and of course, other types of resources, such as CPU resources, memory resources or communication resources. In practical applications, resources such as GPU resources and local storage resources are relatively scarce and costly. Therefore, fragmentation management of these scarce resources according to the resource management method provided in this embodiment can obtain more obvious effects and benefits. Of course, fragmentation management of other types of resources according to the resource management method provided in this embodiment can also obtain effects and benefits. In this embodiment, there is no limitation on the specified type of resources.

[0047] As mentioned above, the resource control method provided in this embodiment can be executed by a control device. In step 200, the control device can monitor the resource status of the specified type of resources and the target type of resources required to be matched on the target physical host. The target physical host can be any physical host that may have the problem of fragmentation of the specified type of resources.

[0048] In this embodiment, the resource status can be used to characterize the usage of the resource. In this embodiment, the characterization dimension used in the resource status is not limited, for example, the occupied amount can be used as the characterization dimension, or the remaining amount can be used as the characterization dimension.

[0049] Among them, for the specified type of resources, there may be one or more resources that need to be matched. For example, for GPU resources, the resources that need to be matched may include CPU resources or memory resources, etc. For the specified type of resources, the type of resources that need to be matched is usually affected by the resource occupation requirements of the relevant instance. It is worth noting that in actual applications, the target type that needs to be paid attention to in step 200 may be pre-specified for the specified type of resources, and the target type may be all resource types or part of the resource types that need to be matched with the specified type of resources. For example, the resources that need to be matched with GPU resources include CPU resources, memory resources, communication resources and network disk resources. In this embodiment, only CPU resources can be used as the target type resource that needs to be paid attention to in step 200. This is because in actual applications, many types of resources on the physical host are usually very sufficient or can be reused in time-sharing. Even if they need to be matched with the specified type of resources, it will not cause the fragmentation problem of the specified type of resources. Therefore, there is no need to pay attention to it.

[0050] In addition, in step 200, various implementations may be used to monitor the resource status of the specified type of resources and the target type of resources required to be matched on the target physical host.

[0051] In an optional implementation method: the target machine model corresponding to the target physical host can be determined; the resource configuration information associated with the target machine model can be queried to determine the total amount of resources corresponding to the specified type of resources and the target type of resources on the target physical host; the occupancy of the specified type of resources and / or the target type of resources by each instance scheduled to the target physical host can be monitored to characterize the resource status.

[0052] In this implementation, the occupied amount is used as the representation dimension of the resource status. Resource configuration information can be recorded separately for different models. Among them, the resource configuration information may include resource types and the total amount of resources corresponding to each type of resource. It should be understood that physical hosts of the same model are consistent in resource configuration. Therefore, after determining the target model corresponding to the target physical host, the control device can query the resource configuration information associated with the target model to determine the total amount of resources corresponding to the specified type of resources and the target type of resources on the target physical host.

[0053] In addition, in this implementation, the control device can perform instance scheduling, that is, the control device can receive an instance creation request and schedule the instance creation request to a suitable physical host to control the hit physical host to respond to the instance creation request and create the required instance. Usually, the required instance specifications are specified in the instance creation request, so that the control device can monitor the instance's occupancy of related resources from the received instance creation request. Accordingly, for the target physical host, the control device can monitor the occupancy of the specified type of resources and / or the target type of resources by each instance scheduled to the target physical host, thereby calculating the corresponding occupied amount of the specified type of resources and the target type of resources on the target physical host.

[0054] In this implementation, the management and control device can accurately monitor the resource status of the specified type of resources and the target type of resources required to match them on the target physical host without the target physical host being aware of it, providing a processing basis for subsequent resource management and control steps.

[0055] It should be understood that the above implementation is exemplary, and other implementations can be used in this embodiment to monitor the resource status of the specified type of resources and the target type of resources required to be matched on the target physical host, but it is not limited to this. For example, the target physical host can actively report the corresponding remaining amount of the specified type of resources and the target type of resources to the control device, so that the remaining amount can be used as the representation dimension of the resource status.

[0056] Continue to refer Figure 2 In step 201, it can be determined whether the resource status monitored in step 200 meets the specification guarantee requirements set for the specified type of resources. If not, a target instance is selected from the instances occupying the target type of resources on the target physical host.

[0057] Here, it is proposed to use the resource status monitored in step 200 as a reference to evaluate whether there is currently a fragmentation problem of the specified type of resources on the target physical host.

[0058] Considering that the essence of the fragmentation problem is the problem of unreasonable resource surplus, this embodiment proposes that in step 201, the surplus of the specified type of resources and the surplus of the target type of resources can be determined according to the respective resource states of the specified type of resources and the target type of resources; if the surplus of the specified type of resources and the surplus of the target type of resources do not meet the specification guarantee requirements, it is determined that the monitored resource state does not meet the specification guarantee requirements. Among them, the specification guarantee requirements may include the specification relationship specified for the specified type of resources and the target type of resources. The specification relationship in this embodiment can be characterized by the ratio between the resource amount of the specified type of resources and the resource amount of the target type of resources, or can be characterized by the numerical values ​​corresponding to the resource amount of the specified type of resources and the resource amount of the target type of resources.

[0059] In step 201, the specification guarantee requirements corresponding to the specified type of resources can be set on the physical host containing the specified type of resources, and the specification guarantee requirements set for the specified type of resources on different physical hosts may not be completely the same. The specification guarantee requirements set for the specified type of resources are affected by the instance types supported by the target physical host. To this end, in an exemplary setting scheme provided in this embodiment: the resource specification information required for each type of instance supported by the target physical host can be referred to to set appropriate specification guarantee requirements for the specified type of resources.

[0060] Specifically, in this exemplary setting scheme: the resource specification information required for each type of instance that needs to occupy a specified type of resource and is supported for deployment on the target physical host can be obtained; from the obtained resource specification information, the specification relationship between the required specified type of resource and the target type of resource is parsed respectively; from the parsed specification relationship, the target specification relationship is selected to set the specification guarantee requirements corresponding to the specified type of resource.

[0061] In this exemplary setting scheme, resource specification information is maintained for each type of instance. Among them, the resource specification information can record the resource type required by the instance and the amount of resources required under each type of resource. In addition, the instance types supported for deployment can also be maintained for each physical host. In this way, the control device can first determine which types of instances supported for deployment on the target physical host need to occupy the specified type of resources, and then query the resource specification information required for each of these types of instances, and parse the specification relationship between the specified type of resources and the target type of resources required by each of these types of instances based on the amount of resources required for each of the specified type of resources and the target type of resources recorded in the resource specification information.

[0062] On this basis, in this exemplary setting scheme, the control device can select a suitable target specification relationship from the specification relationship analyzed above to set the specification guarantee requirements corresponding to the specified type of resources. Preferably, the resource quantity ratio between the specified type of resources and the target type of resources can be used to characterize the specification relationship. Based on this, the minimum value can be selected as the target ratio. Accordingly, the specification guarantee requirement can be designed so that the remaining quantity ratio must be less than or equal to the target ratio.

[0063] For example, the A-type instances, B-type instances, and C-type instances supported for deployment on the target physical server all need to occupy GPU resources. By querying the resource specification information required by each of the three types of instances, it can be analyzed that the specification relationship required by the A-type instance is GPU:CPU=1:8; the specification relationship required by the B-type instance is GPU:CPU=1:4; and the specification relationship required by the C-type instance is GPU:CPU=1:16. On this basis, the GPU:CPU=1:16 required by the C-type instance among the three types of instances can be used as the target specification relationship, and the specification guarantee requirement is designed to be less than or equal to 1:16. The target ratio selected by this strategy and the corresponding specification guarantee requirement can enable the remaining resources on the target physical host to support the construction of various types of instances that need to occupy the specified type of resources deployed by the target physical host. In particular, the instance that requires the largest amount of resources under the target type of resources can also be built on the target physical host in a timely manner, and there will no longer be a problem of being unable to be built in a timely manner due to insufficient resources.

[0064] During the research process, the inventors found that according to the specification guarantee requirements set in the above-mentioned preferred method, when managing and controlling the specified type of resources on the target physical host, a very good fragmentation management effect can be obtained, and fragmentation problems will hardly occur. However, it should be understood that in this exemplary design scheme, other values ​​can also be selected from the resource quantity ratios analyzed above as target ratios, for example, the middle value, the minimum value, etc., which are not limited here. It’s just that when different values ​​are selected, there are slight differences in the fragmentation management effects that can be obtained, but they can all improve the fragmentation problem of the specified type of resources on the target physical host.

[0065] In addition, it is worth noting that in this embodiment, other design solutions can be used to set the specification guarantee requirements for the specified type of resources on the target physical host, and are not limited to the above exemplary setting solutions. For example, a ratio threshold can be set for the specified type of resources in the specification guarantee requirements based on experience, and the specification guarantee requirements can be designed so that the remaining amount ratio monitored in step 200 is less than or equal to the ratio threshold, etc. No more examples of design solutions are given here.

[0066] When it is determined that the resource status monitored in step 200 does not meet the specification guarantee requirement based on the set specification guarantee requirement, a target instance can be selected from the instances occupying the target type resources on the target physical host according to the preset migration strategy for migration.

[0067] In step 201, considering that migrating instances that do not occupy target type resources will not bring any resource supplement effect to target type resources in the target physical host, it is proposed to select a target instance from instances that occupy target type resources.

[0068] The migration strategy that can be adopted in step 201 will be described in detail later, and this embodiment does not limit this.

[0069] Continue to refer Figure 2 In step 202, the target instance may be migrated from the target physical host so that the remaining amount of the target type resources in the target physical host meets the specification guarantee requirement. In this embodiment, after the target instance is migrated out, a part of the target type resources may be released on the target physical host to realize the replenishment of the target type resources on the target physical host, so that more target type resources may appear on the target physical host to be used for matching with the designated type resources, so as to reduce the amount of fragmented designated type resources on the target physical host.

[0070] In this embodiment, the migration destination of the target instance is not limited. In order to avoid the fragmentation of the specified type of resources on other physical hosts caused by the migration of the target instance, in a preferred solution: if the target instance does not occupy the specified type of resources, a physical host that does not support the deployment of instances that need to occupy the specified type of resources can be selected from other physical hosts as the destination host of the target instance; the target instance is migrated to the selected destination host.

[0071] As mentioned above, different physical hosts support different instance types for deployment. Therefore, the control device can query the instance type list supported by each physical host and select a physical host that does not support the deployment of instances that occupy the specified type of resources from other physical hosts other than the target physical host as the destination host selected for the target instance. In this way, the destination host will not have the problem of fragmentation of the specified type of resources due to the migration of the target instance, and since the target instance does not need to occupy the specified type of resources, the failure of the destination host to support the occupation of the specified type of resources will not affect the migration of the target instance.

[0072] This preferred solution can avoid introducing the fragmentation problem of the specified type of resources on the destination host while ensuring that the target instance can be migrated to the destination host without any obstacles. Therefore, the resource utilization rate on the target physical host and the destination host can be effectively guaranteed.

[0073] Of course, in this embodiment, other technical solutions can also be used to select a destination host for the target instance. It is not limited to the above-mentioned preferred solution. For example, if the target instance also needs to occupy a specified type of resource, then a migration prediction can be performed on other physical hosts based on the amount of resources that the target instance needs to occupy under the specified type of resources and the target type of resources, respectively, to predict whether the ratio of the remaining amount between the specified type of resources and the target type of resources on the corresponding physical host after the target instance is migrated to the corresponding physical host still meets the specification guarantee requirements set for the specified type of resources; based on this, a physical host whose prediction result is that the ratio of the remaining amount between the specified type of resources and the target type of resources after the target instance is migrated to the corresponding physical host still meets the corresponding specification guarantee requirements can be selected from other physical hosts as the destination host. This selection method can also avoid the introduction of fragmentation problems for the specified type of resources on the destination host due to the migration of the target instance. No further examples of the selection scheme are given here.

[0074] In summary, in this embodiment, the resource status of the specified type of resources and the target type of resources required to be matched on the target physical host can be monitored separately; if the monitored resource status does not meet the specification guarantee requirements set for the specified type of resources, the target instance is selected from the instances occupying the target type of resources on the target physical host; the target instance is migrated from the target physical host so that the resource status of the specified type of resources and the target type of resources in the target physical host meet the specification guarantee requirements. In this way, the target type of resources on the target physical host can be replenished in a timely manner through resource management and control, avoiding the situation where the remaining amount of the specified type of resources is sufficient, while the remaining amount of the target type of resources required to be matched is insufficient, resulting in the specified type of resources having to be idle, thereby effectively improving the utilization rate of the specified type of resources and ensuring that various instances that need to occupy the specified type of resources can be created more timely.

[0075] In the above or below embodiments, a variety of migration strategies may be used to select a target instance. Figure 3 A schematic diagram of a resource management method provided by another exemplary embodiment of the present application. Figure 3 , the resource management method may include:

[0076] Step 300: monitor the resource status of the specified type of resources and the target type of resources required to be matched on the target physical host;

[0077] Step 301: If the monitored resource status does not meet the specification guarantee requirement set for the specified type of resource, a target instance is selected from instances on the target physical host that occupy the target type of resource and do not occupy the specified type of resource;

[0078] Step 302: Migrate the target instance from the target physical host so that the resource states of the designated type resources and the target type resources in the target physical host meet the specification guarantee requirements.

[0079] Regarding step 300 and step 302, reference may be made to the relevant description in the above-mentioned embodiment, and no repetition is made here. Based on step 301, this embodiment proposes an optional implementation method of selecting a target instance.

[0080] Referring to step 301, this embodiment proposes that a target instance can be selected from instances that occupy target type resources on the target physical host and do not occupy specified type resources. As mentioned in the full text, the control device in this embodiment can complex instance scheduling work, so the control device can obtain information such as the type and specification of the instances running on each physical host. In this way, in step 301, the control device can unimpededly recognize which instances of the instances running on the target physical host are instances that occupy target type resources but do not occupy specified type resources.

[0081] Here, setting the screening range to "instances that occupy target type resources and do not occupy specified type resources" can release a portion of target type resources after the screened target instances are migrated out, thereby increasing the remaining amount of target type resources in the target physical host. Migrating the target instances will not cause the remaining amount of specified type resources on the target physical host to change. Furthermore, this can more efficiently ensure that the remaining ratio between specified type resources and target type resources on the target physical host meets the specification guarantee requirements.

[0082] Under this screening range, an exemplary solution for screening out target instances from this range may be:

[0083] Determine the remaining amount of the target type resource according to the resource status of the target type resource;

[0084] Based on the specification guarantee requirements and the remaining amount of target type resources, calculate the freed-up amount of target type resources required to meet the specification guarantee requirements;

[0085] With the goal of reaching the freed-up amount for the total amount of occupied resources of the target type, a target instance is selected from instances on the target physical host that occupy resources of the target type and do not occupy resources of the specified type.

[0086] In this exemplary solution, since the instances within the screening range will not cause the remaining amount of the specified type of resource to change, we only need to focus on whether the remaining amount of the target type of resource caused by the selected target instance can meet the required specification guarantee requirements. For this reason, the parameter of the freed-up amount of the target type of resource is introduced in this exemplary solution as a basis for screening the target instance. On this basis, we can screen out instances within the aforementioned screening range whose total occupation of the target type of resource can reach the freed-up amount as the target server.

[0087] In actual applications, the number of the selected target servers may be one or more. When the number of target servers is one, the target server's occupation of the target type resources on the target physical server has reached the aforementioned required freed-up amount, so that after the target server is migrated out, the amount of target type resources freed up is added to the remaining amount of target type resources already on the target physical host, and the remaining amount ratio between the specified type resources and the target type resources can meet the required specification guarantee requirements. When the number of target physical servers is multiple, the total amount of target type resources occupied by multiple target servers on the target physical server can reach the aforementioned required freed-up amount, so that after the target servers are migrated out, the amount of target type resources freed up is added to the remaining amount of target type resources already on the target physical host, and the remaining amount ratio between the specified type resources and the target type resources can meet the required specification guarantee requirements.

[0088] It is worth noting that, in this embodiment, in addition to introducing the parameter of freed-up amount as a screening basis as in the aforementioned exemplary solution, other screening criteria may also be used, for example, the total amount of target type resources occupied by the remaining instances on the target physical host after excluding the target instance, etc. No further examples are given here.

[0089] The inventors found during the research that there are usually multiple selectable instances within the aforementioned screening range, so how to reasonably select the target instance becomes a problem that needs to be considered. Two exemplary screening methods are provided below.

[0090] The first one:

[0091] The instances occupying the target type resources and not occupying the specified type resources on the target physical host can be sorted according to the occupation of the target type resources;

[0092] Select instances in order of occupancy from low to high until the selected instances make the remaining amount of target type resources in the target physical host meet the required specification guarantee requirements;

[0093] Use the selected instance as the target instance.

[0094] In this screening method, instances with lower occupancy of target type resources can be preferably selected from the screening range. After a sufficient number of instances are screened out, the screening is stopped, so that the screening result can be obtained. The sufficient amount here means that the freed-up amount caused by the selected instances or other parameters used as the screening basis in this embodiment have met the relevant requirements. For example, the freed-up amount caused by the selected instances is equal to or greater than the required freed-up amount mentioned above.

[0095] In this way, the target instances screened out by this screening method will mostly be instances of relatively small specifications, which can effectively ensure the migration success rate and more efficiently achieve the management of fragmented resources.

[0096] Second type:

[0097] The instances that occupy the target type of resources and do not occupy the specified type of resources on the target physical host can be sorted according to the current load;

[0098] Instances are selected in order from low to high load, and the selected instances ensure that the remaining amount of target type resources in the target physical host meets the required specification guarantee requirements;

[0099] Use the selected instance as the target instance.

[0100] In this screening method, instances with lower load can be selected from the screening range. After a sufficient number of instances are screened, the screening is stopped, and the screening results can be obtained. In this way, the target instances screened in this screening method will mostly be instances with relatively small loads, the amount of data involved in the migration will be relatively small, and the migration efficiency will be higher. Therefore, the management of fragmented resources can be achieved more efficiently.

[0101] It should be understood that the above two screening methods are only exemplary. In this embodiment, other screening methods may also be used, such as random screening or preferentially selecting instances with a later creation time, etc. No further examples are given here.

[0102] In summary, in this embodiment, an optional implementation method for selecting a target instance is provided based on step 301. In this optional implementation method, the screening range is limited to instances that occupy target type resources but do not occupy specified type resources, and the target instance can be screened out within the screening range to ensure that the remaining amount of target type resources in the target physical host can meet the required specification guarantee requirements after the target instance is migrated out. Under the limitation of the screening range, the selected target instance can supplement the target type resources on the target physical host after being migrated out from the target physical host, but it will not cause the remaining amount of specified type resources on the target physical host to change. Therefore, this can effectively improve the efficiency of fragmentation management of specified type resources.

[0103] It is worth noting that in addition to the optional implementation method of selecting the target instance provided in step 301, other implementation methods can also be used to select the target instance in this embodiment. For example, the screening range is set to instances occupying resources of the target type, etc. No further implementation examples are given here.

[0104] In the above or following embodiments, the management and control device does not need to perform resource management on all physical hosts, but can screen out physical hosts that may have fragmentation problems from the deployed physical hosts, and perform fragmentation processing on them according to the resource management method provided in the above embodiments. Figure 4 A schematic diagram of a resource management method provided by another exemplary embodiment of the present application. Figure 4 , the method may include:

[0105] Step 400: in the case of a newly deployed physical host, determine the target machine model corresponding to the newly deployed physical host;

[0106] Step 401: query the resource configuration information associated with the target model to determine whether a specified type of resource exists on the newly deployed physical host;

[0107] Step 402: query the instance type list supported by the newly deployed physical host to determine whether the newly deployed physical host supports mixed deployment of multiple types of instances;

[0108] Step 403: If the newly deployed physical host has resources of the specified type and supports mixed deployment of multiple types of instances, the newly deployed physical host is determined as the target physical host;

[0109] Step 404: if the remaining amount ratio between the specified type of resource on the monitored target physical host and the target type of resource required to be matched does not meet the specification guarantee requirement set for the specified type of resource, then select a target instance from the instances occupying the target type of resource on the target physical host;

[0110] Step 405: Migrate the target instance from the target physical host so that the remaining amount of the target type resources in the target physical host meets the specification guarantee requirement.

[0111] Among them, step 404 and step 405 can refer to the relevant description in the above-mentioned embodiment, and will not be repeated here.

[0112] In this embodiment, the control device is aware of the deployment and offline of the physical host. Therefore, when a newly deployed physical host appears, the control device can determine whether the physical host may have a fragmentation problem corresponding to the specified type of resources according to steps 400 to 403 in this embodiment.

[0113] refer to Figure 4 In step 400, the target machine model corresponding to the newly deployed physical host can be determined. As mentioned above, different machine models are associated with their own resource configuration information. Based on this, in step 401, the resource configuration information associated with the target machine model can be queried to determine whether there are specified types of resources on the newly deployed physical host.

[0114] refer to Figure 4 In this embodiment, it is further proposed in step 402 that a list of instance types supported by the newly deployed physical host can be queried to determine whether the newly deployed physical host supports mixed deployment of multiple types of instances. For a physical host that does not support mixed deployment of multiple types of instances, since the instance type carried on it is single, the ratio of the remaining amount between each type of resource is usually relatively stable, and therefore, fragmented resources usually do not occur. On a physical host that supports mixed deployment, since different types of instances require different types of resources and resource specifications, the creation or release of instances on such physical hosts will cause the ratio of the remaining amount between each type of resources to change dynamically, and therefore, fragmented resources often occur.

[0115] Accordingly, in step 403, if the newly deployed physical host has resources of the specified type and supports mixed deployment of multiple types of instances, the newly deployed physical host is determined as the target physical host.

[0116] Thereafter, the control device can perform real-time resource control on each determined target physical host according to the resource control scheme provided in step 404-step 405, so as to perform fragmentation management on the specified type of resources on these target physical hosts. It should be understood that for the control device, the target physical host determined by it can be one or more, and with the deployment and offline of the physical host, the target physical host concerned by the control device is also dynamically changed. The newly deployed physical host can be added as the target physical host, and the offline target physical host will no longer be concerned.

[0117] In summary, in this embodiment, physical hosts that have resources of a specified type and support mixed deployment of multiple types of instances can be selected from the deployed physical hosts as target physical hosts, and resource management and control can be performed on each of the selected physical hosts to perform fragmentation management on the resources of the specified type on them. This can comprehensively and accurately improve the fragmentation problem of various resources in the deployed physical hosts and improve the resource utilization rate in the deployed physical hosts.

[0118] Figure 5a and 5b This is a schematic diagram of an exemplary application scenario provided by this application. Figure 5a and Figure 5b , taking the fragmentation management of GPU resources in physical host A as an example, a detailed description of the memory management solution is given.

[0119] refer to Figure 5aIn this exemplary application scenario, physical host A can provide 4 GPUs and 64 CPU cores, and physical host A supports hybrid deployment. After the scheduling of the management and control equipment, physical host A first created a GPU instance (actually an instance) according to customer demand 1, and the GPU instance occupied 2 GPUs and 8 CPU cores. Afterwards, a CPU instance (actually also an instance) was created according to customer demand 2, and the CPU instance occupied 40 CPU cores. At this point, there are 2 GPUs and 16 CPU cores left on physical host A. But after a while, the aforementioned customer demand 1 released 2 GPUs and 8 CPU cores.

[0120] Afterwards, when the control device monitors the resources of physical host A, it can be detected that the remaining ratio between the GPU and the CPU on physical host A is: 4:24, and the control device also obtains the specification guarantee requirement set for the GPU of physical host A: the remaining ratio between the GPU and the CPU is less than or equal to 1:8. Therefore, the control device detects that a GPU fragmentation problem has occurred on physical host A at this time.

[0121] refer to Figure 5b , the control device can filter out some CPU instances from physical host A and migrate these CPU instances to other physical hosts to free up more CPU cores on physical host A, so that the GPU:CPU on physical host A can meet the 1:8 requirement.

[0122] After the above memory management process, the remaining ratio between GPU and CPU on physical host A can be kept stable at about 1:8 without large fluctuations. In this way, it can be ensured that physical host A can take over instances with specifications of GPU:CPU ≥ 1:8 at any time, especially ensuring the timely deployment of instances with the maximum specification of GPU:CPU = 1:8, thereby improving the resource utilization of physical host A.

[0123] Figure 6 This is a schematic diagram of a control device provided by an exemplary embodiment of the present application. Figure 6 As shown, the management and control device includes: a memory 60 and a processor 61.

[0124] The processor 61 is coupled to the memory 60 and is configured to execute the computer program in the memory 60 to:

[0125] Monitor the resource status of the specified type of resources and the target type of resources they need to match on the target physical host respectively;

[0126] If the monitored resource status does not meet the specification guarantee requirement set for the specified type of resource, selecting a target instance from the instances occupying the target type of resource on the target physical host;

[0127] The target instance is migrated from the target physical host so that the resource states of the designated type of resources and the target type of resources in the target physical host meet the specification guarantee requirement.

[0128] In an optional embodiment, the processor 61 may be configured to:

[0129] Determining the remaining amount of the designated type of resource and the remaining amount of the target type of resource according to the respective resource states of the designated type of resource and the target type of resource;

[0130] If the remaining amount of the designated type of resource and the remaining amount of the target type of resource do not meet the specification guarantee requirement, determining that the monitored resource state does not meet the specification guarantee requirement;

[0131] The specification guarantee requirement includes the specification relationship specified for the specified type of resources and the target type of resources.

[0132] In an optional embodiment, when the processor 61 selects the target instance from the instances occupying the target type resource on the target physical host, it can be used to:

[0133] The target instance is selected from instances on the target physical host that occupy resources of the target type and do not occupy resources of the specified type.

[0134] In an optional embodiment, when the processor 61 selects the target instance from instances on the target physical host that occupy the target type of resources and do not occupy the specified type of resources, it may be configured to:

[0135] Determining the remaining amount of the target type resource according to the resource status of the target type resource;

[0136] Based on the specification guarantee requirement and the remaining amount of the target type resource, calculating the freed-up amount of the target type resource required to meet the specification guarantee requirement;

[0137] With the total amount of occupation of the target type of resources reaching the vacated amount as a goal, the target instance is selected from instances on the target physical host that occupy the target type of resources and do not occupy the designated type of resources;

[0138] The number of the target instances is one or more.

[0139] In an optional embodiment, when selecting a target instance, the processor 61 may be configured to:

[0140] According to the occupation amount of the target type resource, the instances on the target physical host that occupy the target type resource and do not occupy the designated type resource are sorted;

[0141] Selecting instances in order of occupancy from low to high, until the selected instances make the remaining amount of the target type resources in the target physical host meet the specification guarantee requirement;

[0142] The selected instance is used as the target instance.

[0143] In an optional embodiment, when selecting a target instance, the processor 61 may be configured to:

[0144] According to the current load, the instances on the target physical host that occupy the target type of resources and do not occupy the specified type of resources are sorted;

[0145] Selecting instances in order of load from low to high until the selected instances make the remaining amount of the target type resources in the target physical host meet the specification guarantee requirement;

[0146] The selected instance is used as the target instance.

[0147] In an optional embodiment, when the processor 61 monitors the resource status of the specified type of resource and the target type of resource required to be matched on the target physical host respectively, it can be used to:

[0148] Determine a target machine model corresponding to the target physical host;

[0149] Querying the resource configuration information associated with the target machine model to determine the total amount of resources corresponding to the designated type of resources and the target type of resources on the target physical host;

[0150] The occupation of the designated type of resources and / or the target type of resources by each instance scheduled to the target physical host is monitored to characterize the resource status.

[0151] In an optional embodiment, the processor 61 may also be configured to:

[0152] In the case of a newly deployed physical host, determine the target machine model corresponding to the newly deployed physical host;

[0153] Querying resource configuration information associated with the target machine model to determine whether the specified type of resource exists on the newly deployed physical host;

[0154] Querying a list of instance types supported by the newly deployed physical host to determine whether the newly deployed physical host supports mixed deployment of multiple types of instances;

[0155] If the designated type of resource exists on the newly deployed physical host and supports mixed deployment of multiple types of instances, the newly deployed physical host is determined as the target physical host.

[0156] In an optional embodiment, when migrating the target instance from the target physical host, the processor 61 may be configured to:

[0157] If the target instance does not occupy the specified type of resources, a physical host that does not support the deployment of instances that need to occupy the specified type of resources is selected from other physical hosts as the destination host of the target instance;

[0158] Migrate the target instance to the selected destination host.

[0159] In an optional embodiment, the processor 61 may also be configured to:

[0160] Obtain resource specification information required by each type of instance that needs to occupy the specified type of resources and is supported for deployment on the target physical host;

[0161] From the acquired resource specification information, respectively parse out the specification relationship between the required specified type resources and the target type resources;

[0162] A target specification relationship is selected from the parsed specification relationships to set specification assurance requirements corresponding to the specified type of resources.

[0163] In an optional embodiment, the specified type of resources include one or more resources of GPU resources or local storage resources.

[0164] Further, if Figure 6 As shown, the control device also includes: a communication component 62 and a power supply component 63 and other components. Figure 6 Only some components are shown schematically, which does not mean that the control equipment only includes Figure 6 Components shown.

[0165] It is worth noting that the technical details in the above-mentioned embodiments of the control device can be referred to the relevant description in the aforementioned method embodiment. In order to save space, they will not be repeated here, but this should not cause a loss in the scope of protection of this application.

[0166] In addition to the resource control method implemented by the control device mentioned in the above embodiment, the physical host can also independently perform resource fragmentation management. Figure 7 A schematic diagram of the structure of a physical host provided by another exemplary embodiment of the present application. Figure 6As shown, the physical host may include: a memory 70 and a processor 71. The physical host provides designated type resources and target type resources and carries one or more instances.

[0167] The processor 71 is coupled to the memory 70 and is configured to execute the computer program in the memory 70 to:

[0168] Respectively monitoring the resource status of the designated type of resource and the target type of resource required to be matched therewith;

[0169] If the monitored resource status does not meet the specification guarantee requirement set for the specified type of resource, a target instance is selected from instances occupying the target type of resource;

[0170] The target instance is migrated out so that the resource states of the designated type resources and the target type resources meet the specification guarantee requirement.

[0171] In this embodiment, the physical host can autonomously perform fragmentation management of designated types of resources.

[0172] In an optional embodiment, the processor 71 may also be configured to:

[0173] Determining the remaining amount of the designated type of resource and the remaining amount of the target type of resource according to the respective resource states of the designated type of resource and the target type of resource;

[0174] If the remaining amount of the designated type of resource and the remaining amount of the target type of resource do not meet the specification guarantee requirement, determining that the monitored resource state does not meet the specification guarantee requirement;

[0175] The specification guarantee requirement includes the specification relationship specified for the specified type of resources and the target type of resources.

[0176] In an optional embodiment, when the processor 71 selects the target instance from the instances occupying the target type resource, it can be used to:

[0177] The target instance is selected from instances that occupy resources of the target type and do not occupy resources of the specified type.

[0178] In an optional embodiment, when the processor 71 selects the target instance from instances that occupy the target type of resources and do not occupy the specified type of resources, it may be configured to:

[0179] Determining the remaining amount of the target type resource according to the resource status of the target type resource;

[0180] Based on the specification guarantee requirement and the remaining amount of the target type resource, calculating the freed-up amount of the target type resource required to meet the specification guarantee requirement;

[0181] With the total amount of occupation of the target type resources reaching the vacated amount as a goal, the target instance is selected from instances that occupy the target type resources and do not occupy the designated type resources;

[0182] The number of the target instances is one or more.

[0183] In an optional embodiment, when selecting a target instance, the processor 71 may be configured to:

[0184] According to the occupation amount of the target type resource, the instances that occupy the target type resource and do not occupy the specified type resource are sorted;

[0185] Select instances in order of occupancy from low to high, until the selected instances make the remaining amount of the target type resources meet the specification guarantee requirement;

[0186] The selected instance is used as the target instance.

[0187] In an optional embodiment, when selecting a target instance, the processor 71 may be configured to:

[0188] According to the current load, the instances that occupy the target type of resources and do not occupy the specified type of resources are sorted;

[0189] Select instances in order of load from low to high until the selected instances make the remaining amount of the target type resources meet the specification guarantee requirement;

[0190] The selected instance is used as the target instance.

[0191] In an optional embodiment, when the processor 71 monitors the resource status of the specified type resource and the target type resource required to be matched respectively, it can be used to:

[0192] Determine the target model that corresponds to itself;

[0193] Querying resource configuration information associated with the target machine model to determine the total amount of resources corresponding to the designated type of resources and the target type of resources;

[0194] Monitor the occupation of the designated type of resources and / or the target type of resources by each instance carried by the system to characterize the resource status.

[0195] In an optional embodiment, when migrating the target instance, the processor 71 may be used to:

[0196] If the target instance does not occupy the specified type of resources, a physical host that does not support the deployment of instances that need to occupy the specified type of resources is selected from other physical hosts as the destination host of the target instance;

[0197] Migrate the target instance to the selected destination host.

[0198] In an optional embodiment, the processor 71 may also be configured to:

[0199] Obtain resource specification information required by each type of instance that needs to occupy the specified type of resources and that is supported for deployment;

[0200] From the acquired resource specification information, respectively parse out the specification relationship between the required specified type resources and the target type resources;

[0201] A target specification relationship is selected from the parsed specification relationships to set specification assurance requirements corresponding to the specified type of resources.

[0202] In an optional embodiment, the specified type of resources include one or more resources of GPU resources or local storage resources.

[0203] Further, if Figure 7 As shown, the physical host also includes other components such as a communication component 72 and a power supply component 73. Figure 7 Only some components are shown schematically, which does not mean that the physical host only includes Figure 7 Components shown.

[0204] Accordingly, an embodiment of the present application also provides a computer-readable storage medium storing a computer program, which, when executed, can implement the steps in the above method embodiment that can be executed by a control device or a physical host.

[0205] Above Figure 6 and Figure 7 The memory in the computer is used to store computer programs and can be configured to store various other data to support operations on the computing platform. Examples of such data include instructions for any application or method operating on the computing platform, contact data, phone book data, messages, pictures, videos, etc. The memory can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0206] Above Figure 6 and Figure 7 The communication component in is configured to facilitate wired or wireless communication between the device where the communication component is located and other devices. The device where the communication component is located can access a wireless network based on a communication standard, such as WiFi, 2G, 3G, 4G / LTE, 5G and other mobile communication networks, or a combination thereof. In an exemplary embodiment, the communication component receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

[0207] Above Figure 6 and Figure 7 The power supply component in the device provides power to various components of the device where the power supply component is located. The power supply component may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the device where the power supply component is located.

[0208] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0209] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0210] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0211] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0212] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0213] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0214] The above is only the embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A resource management method, characterized in that: include: Monitor the resource status of the specified type of resources and the target type of resources required to match them on the target physical host respectively; If the monitored resource status does not meet the specification guarantee requirement set for the specified type of resource, selecting a target instance from the instances occupying the target type of resource on the target physical host; The target instance is migrated from the target physical host so that the resource states of the designated type of resources and the target type of resources in the target physical host meet the specification guarantee requirement.

2. The method according to claim 1, characterized in that Also includes: Determining the remaining amount of the designated type of resource and the remaining amount of the target type of resource according to the respective resource states of the designated type of resource and the target type of resource; If the remaining amount of the designated type of resource and the remaining amount of the target type of resource do not meet the specification guarantee requirement, determining that the monitored resource state does not meet the specification guarantee requirement; The specification guarantee requirement includes the specification relationship specified for the specified type of resources and the target type of resources.

3. The method according to claim 1, characterized in that Selecting a target instance from instances occupying the target type resource on the target physical host includes: The target instance is selected from instances on the target physical host that occupy resources of the target type and do not occupy resources of the specified type.

4. The method according to claim 3, characterized in that: Selecting the target instance from instances on the target physical host that occupy resources of the target type and do not occupy resources of the specified type includes: Determining the remaining amount of the target type resource according to the resource status of the target type resource; Based on the specification guarantee requirement and the remaining amount of the target type resource, calculating the freed-up amount of the target type resource required to meet the specification guarantee requirement; With the total amount of occupation of the target type of resources reaching the vacated amount as a goal, the target instance is selected from instances on the target physical host that occupy the target type of resources and do not occupy the designated type of resources; The number of the target instances is one or more.

5. The method according to any one of claims 3 or 4, characterized in that: Selecting the target instance includes: According to the occupation amount of the target type resource, the instances on the target physical host that occupy the target type resource and do not occupy the designated type resource are sorted; Selecting instances in order of occupancy from low to high, until the selected instances make the remaining amount of the target type resources in the target physical host meet the specification guarantee requirement; The selected instance is used as the target instance.

6. The method according to any one of claims 3 or 4, characterized in that: Selecting the target instance includes: According to the current load, the instances on the target physical host that occupy the target type of resources and do not occupy the specified type of resources are sorted; Selecting instances in order of load from low to high until the selected instances make the remaining amount of the target type resources in the target physical host meet the specification guarantee requirement; The selected instance is used as the target instance.

7. The method according to claim 1, characterized in that Monitor the resource status of the specified type of resources and the target type of resources they need to match on the target physical host, including: Determine a target machine model corresponding to the target physical host; Querying the resource configuration information associated with the target machine model to determine the total amount of resources corresponding to the designated type of resources and the target type of resources on the target physical host; The occupation of the designated type of resources and / or the target type of resources by each instance scheduled to the target physical host is monitored to characterize the resource status.

8. The method according to claim 1, characterized in that: Also includes: In the case of a newly deployed physical host, determine the target machine model corresponding to the newly deployed physical host; Querying resource configuration information associated with the target machine model to determine whether the specified type of resource exists on the newly deployed physical host; Querying a list of instance types supported by the newly deployed physical host to determine whether the newly deployed physical host supports mixed deployment of multiple types of instances; If the designated type of resource exists on the newly deployed physical host and supports mixed deployment of multiple types of instances, the newly deployed physical host is determined as the target physical host.

9. The method according to claim 1, characterized in that: Migrating the target instance from the target physical host includes: If the target instance does not occupy the specified type of resources, a physical host that does not support the deployment of instances that need to occupy the specified type of resources is selected from other physical hosts as the destination host of the target instance; Migrate the target instance to the selected destination host.

10. The method according to any one of claims 1 or 2, characterized in that: Also includes: Obtain resource specification information required by each type of instance that needs to occupy the specified type of resources and is supported for deployment on the target physical host; From the acquired resource specification information, respectively parse out the specification relationship between the required specified type resources and the target type resources; A target specification relationship is selected from the parsed specification relationships to set specification assurance requirements corresponding to the specified type of resources.

11. The method according to claim 1, characterized in that: The specified type of resources includes one or more resources of GPU resources or local storage resources.

12. A control device, characterized in that: Includes memory, processor, and communication components; The memory is used to store one or more computer instructions; The processor is coupled to the memory and the communication component, and is used to execute the one or more computer instructions to execute the resource management method described in any one of claims 1-11.

13. A physical host, characterized in that: The physical host comprises a memory, a processor and a communication component, wherein the physical host provides resources of a specified type and resources of a target type and carries one or more instances; The memory is used to store one or more computer instructions; The processor is coupled to the memory and the communication component and is configured to execute the one or more computer instructions for: Respectively monitoring the resource status of the designated type of resource and the target type of resource required to be matched therewith; If the monitored resource status does not meet the specification guarantee requirement set for the specified type of resource, a target instance is selected from instances occupying the target type of resource; The target instance is migrated out so that the resource states of the designated type resources and the target type resources meet the specification guarantee requirement.

14. A computer-readable storage medium storing computer instructions, characterized in that: When the computer instructions are executed by one or more processors, the one or more processors are caused to execute the resource management method according to any one of claims 1 to 10.