Method for providing cloud service, cloud platform, and related apparatus

Through the Elastic Memory Accelerated Service (EMAS) management node, tenants allow custom cloud service resources, solving problems that are difficult to meet in the existing technology, improving resource utilization and performance, and reducing operation and maintenance costs.

WO2024245394A1PCT designated stage expired Publication Date: 2024-12-05HUAWEI CLOUD COMPUTING TECHNOLOGIES CO LTD
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
PCT/CN2024/096654
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-08-21
Filing Date
2024-05-31
Publication Date
2024-12-05

AI Technical Summary

Technical Problem

In existing cloud computing technology, the memory capacity requirements of tenants are difficult to meet, and the preset specification limitations lead to low resource utilization and high operation and maintenance costs, and the inability to effectively utilize idle resources in the data center.

Method used

Provide a cloud service method to manage nodes through Elastic Memory Acceleration Service (EMAS), allowing tenants to customize computing resources and memory resources according to business needs, create Elastic Memory Acceleration Super Instance (EMA Super Instance), and obtain the tenant's Attribute information, dynamically adjust resource nodes to meet demand.

Benefits of technology

It improves the performance and resource utilization of cloud services, reduces resource waste and operation and maintenance costs, and realizes the full utilization of idle resources in the data center.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN2024096654_05122024_PF_FP_ABST
    Figure CN2024096654_05122024_PF_FP_ABST
Patent Text Reader

Abstract

The present application relates to the technical field of cloud computing, and discloses a method for providing a cloud service, a cloud platform, and a related apparatus. The method comprises: providing a configuration interface; acquiring attribute information of a target cloud service from the configuration interface, and on the basis of the attribute information, creating a target elastic memory acceleration (EMA) super-instance in a target resource node comprised among a plurality of resource nodes; and setting a management interface of the target EMA super-instance on the basis of a providing mode of the target cloud service, and providing the management interface to a tenant. The tenant will not be limited by a preset specification, and can customize the computing resources and the memory resources of the target cloud service according to service requirements, thereby meeting the requirements of the tenant, and improving the performance of the cloud service. In addition, the resources of at least one resource node can be integrated to provide a service for the tenant, so that the user requirements are met while the utilization rate of idle resources of a data center can be effectively improved, and the resource waste and the operations and maintenance costs of a cloud service provider are reduced.
Need to check novelty before this filing date? Find Prior Art

Description

Method for providing cloud services, cloud platform and related devices

[0001] This application claims priority to Chinese patent application No. 202310655416.6, filed on June 2, 2023, with the invention name “Method, cloud platform and related devices for providing cloud services”, and claims priority to Chinese patent application No. 202311054802.6, filed on August 21, 2023, with the invention name “Method, cloud platform and related devices for providing cloud services”, the entire contents of which are incorporated by reference into this application. Technical Field

[0002] The present application relates to the field of cloud computing technology, and in particular to a method for providing cloud services, a cloud platform, and related devices. Background Art

[0003] With the continuous development of information technology, especially the advancement of the internet, large government, enterprises, and financial institutions have begun to build their own data centers to meet their business resource needs. Simply adding physical equipment is extremely expensive. Against this backdrop, cloud computing emerged. Leveraging virtualization technology, cloud service providers integrate resources and provide unified services, significantly improving resource utilization efficiency.

[0004] In related technologies, cloud service providers provide basic resource services to tenants through virtual machines, containers, and other means. Taking virtual machines as an example, the specifications of virtual machines are predefined. That is, tenants need to specify the number of central processing units (CPUs) and memory size of the virtual machine from the preset specifications. If the virtual machine specification specified by the tenant from the preset specifications is 4U8G, this specification means that the virtual machine has 4 CPU cores and 8 gigabytes (GB) of memory. When the tenant uses the virtual machine, the resources cannot exceed the specified specifications.

[0005] However, in cloud computing scenarios, with the development of business and the evolution of applications, tenants have an increasing demand for memory capacity, and the current preset specifications often cannot meet the needs of some tenants.

[0006] Summary of the Invention

[0007] This application provides a method, apparatus, device, storage medium, and computer program for providing cloud services, which can solve the problem that related technologies cannot meet user needs. The technical solution is as follows:

[0008] In a first aspect, a method for providing cloud services is provided, which is applied to an elastic memory acceleration service EMAS management node included in a cloud platform, and the cloud platform also includes multiple resource nodes, and the resource nodes are used to provide at least one of computing resources and memory resources. The method includes: providing a configuration interface, and the configuration interface is used to prompt tenants to input attribute information of the target cloud service to be provided according to business requirements, and the attribute information includes the provision method of the target cloud service and the performance configuration information of the target cloud service; obtaining the attribute information of the target cloud service from the configuration interface; based on the attribute information, creating a target elastic memory acceleration EMA super instance in the target resource nodes included in the multiple resource nodes, and the target EMA super instance is used to provide the target cloud service for the tenant, and the target resource node includes at least one resource node, and the available resources of the target resource node meet the business requirements; based on the provision method of the target cloud service, setting a management interface of the target EMA super instance, and providing the management interface to the tenant.

[0009] Since the attribute information of the target cloud service is input by the tenant according to business needs, that is, the tenant does not need to be restricted by preset specifications and can directly customize the computing resources and memory resources of the target cloud service according to their own business needs, thereby meeting the needs of the tenant and improving the performance of the cloud service. In addition, since the target resource node includes at least one resource node, that is, the present application can integrate the resources of at least one resource node to provide services to the tenant. In this way, while meeting user needs, it can also effectively improve the utilization rate of idle resources in the data center, reduce resource waste and reduce the operation and maintenance costs of the cloud service provider.

[0010] When the target cloud service selected by the tenant is provided in different ways, the performance configuration information of the target cloud service includes different contents, which will be introduced below.

[0011] In the first case, the target cloud service is provided as a super-instance. In this case, the target cloud service's performance configuration information includes usage type, memory capacity, capacity guarantee mode, capacity step, computing resources, and bandwidth. Usage type indicates whether the target EMA super-instance provides services in a mounted form; memory capacity indicates the target EMA super-instance's memory capacity; capacity guarantee mode indicates whether the target EMA super-instance's memory resources are distributed across variable resource nodes; capacity step indicates the memory size of each EMAI included in the target EMA super-instance; computing resources indicates the target EMA super-instance's computing performance; and bandwidth indicates the target EMA super-instance's data transfer volume per unit time.

[0012] Optionally, the usage type includes single instance usage and mounted usage. Single instance usage means that the target EMA super instance does not provide services in a mounted form. Mounted usage means that the target EMA super instance provides services in a mounted form.

[0013] Optionally, the memory capacity is divided into guaranteed capacity and burst capacity, wherein the guaranteed capacity refers to the minimum memory capacity required to be guaranteed by the target EMA super instance, and the burst capacity indicates the maximum capacity of the target EMA super instance, which is greater than or equal to the guaranteed capacity.

[0014] Optionally, capacity guarantee modes include dynamic and static modes. Dynamic mode is a rotational mode, meaning the EMAS management node can split and relocate the EMAIs included in the target EMA super-instance based on the resource usage of the underlying physical server. Static mode is a non-rotational mode, meaning that once an EMASI is created, it cannot be split or relocated.

[0015] Splitting refers to providing resources originally provided by one EMAI through at least two EMAIs. Relocation refers to providing resources originally provided by a first resource node through resource nodes other than the first resource node.

[0016] Optionally, the default capacity guarantee mode of the EMAS management node is dynamic mode, that is, when the tenant does not select the capacity guarantee mode, the default capacity guarantee mode is dynamic mode, so that resources can be scheduled according to the resource usage of the underlying physical server, thereby achieving full utilization of resources.

[0017] Optionally, the capacity step specifies the minimum memory capacity of each EMAI within the target EMA superinstance. That is, the memory capacity of each EMAI must be no less than the capacity step, and the capacity step sets a lower limit on the memory capacity specifications of all EMAIs comprising the EMA superinstance. In this case, the number of EMAIs comprising the EMA superinstance must be no greater than the actual memory capacity of the EMA superinstance divided by the capacity step.

[0018] Optionally, if the tenant sets the capacity step to 0, in this case, it does not mean that the memory resources corresponding to each EMASI are 0, but rather that the EMAS management node can create EMASI for the tenant based on the resource usage of the underlying physical server without being constrained by the capacity step. The size of the EMASI memory resources is related to the size of the resources actually allocated by the EMAS management node.

[0019] Bandwidth may refer to the amount of data transmitted per second by the target EMA superinstance, in GB per second. Of course, bandwidth may also be represented in other forms in practical applications.

[0020] Optionally, the unit of computing resources is a virtual computing unit (VCU). A VCU can be a core in a physical CPU or data processing unit (DPU), or a hyperthread of a core in a physical CPU or DPU. It should be noted that a VCU can be a computing resource used solely by the target EMA super-instance, or a computing resource shared with other EMA super-instances.

[0021] Since usage types include single instance usage and mounted usage, the performance configuration information of the target cloud service may also include some different information in different usage types, which will be introduced below.

[0022] In case 1, the usage type is single instance usage. That is, the usage type indicates that the target EMA super instance does not provide services in a mounted form. In this case, the performance configuration information of the target cloud service also includes a latency specification and a placement strategy. The latency specification indicates the access latency of the target EMA super instance's memory resources, and the placement strategy indicates whether the target EMA super instance's memory resources and computing resources are located on the same resource node.

[0023] Optionally, the latency specification includes performance priority or cost priority. Performance priority means that the EMAS management node selects the resource node with the shortest latency from multiple resource nodes as the target resource node. Cost priority means that the EMAS management node selects the resource node with the lowest price from multiple resource nodes as the target resource node.

[0024] In practice, in addition to directly selecting performance or cost priority, tenants can also enter a specific latency threshold to determine the latency specification. In this case, the EMAS management node determines the target resource node based on the latency threshold, and the access latency of the memory resource of the target EMA super-instance is ultimately determined to be no greater than the latency threshold.

[0025] Optionally, the EMAS management node sets the default latency specification to cost priority, that is, when the tenant does not select a latency specification, the default latency specification is cost priority. Since resources with low utilization rates are usually lower in price, the EMAS management node can schedule lower-priced resources for tenants based on the resource usage of the underlying physical server, thereby fully utilizing the idle resources of the data center.

[0026] Optionally, the placement strategy includes affinity and anti-affinity, wherein affinity means that the memory resources and computing resources of the target EMA superinstance are located in the same resource node, and anti-affinity means that the memory resources and computing resources of the target EMA superinstance are not located in the same resource node.

[0027] It should be noted that if the tenant does not fill in the placement strategy, the EMAS management node can determine the resource node where the computing resources and memory resources of the target EMA super instance are located based on the resource usage of the underlying physical server.

[0028] Case 2: The usage type is mounted usage, that is, the usage type indicates that the target EMA super-instance provides services in a mounted form. In this case, the performance configuration information of the target cloud service also includes the identifier, latency specification, placement strategy, access strategy, and usage mode of at least one target instance. The target instance refers to the instance used to mount the target EMA super-instance. The latency specification indicates the access latency of the target instance to the target EMA super-instance. The placement strategy indicates whether the target EMA super-instance and a first instance are located on the same resource node. The first instance is one of the at least one target instance. The access strategy indicates the read and write permissions of the at least one target instance to the target EMA super-instance. The usage mode indicates the memory addressing method of the target EMA super-instance and a second instance is one of the at least one target instance.

[0029] It should be noted that the target instance can be a virtual machine instance, a bare metal server instance, or a container instance. The target instance needs to be created using an image that supports EMA capabilities.

[0030] Optionally, the latency specification includes performance priority or cost priority. Performance priority means that the EMAS management node selects the resource node with the shortest access latency to the target instance from multiple resource nodes as the target resource node. Cost priority means that the EMAS management node selects the resource node with the lowest cost from multiple resource nodes as the target resource node.

[0031] In practice, in addition to directly selecting performance or cost priority, tenants can also enter a specific latency threshold to determine the latency specification. In this case, the EMAS management node determines the target resource node based on the latency threshold, ensuring that the target instance's access latency to the target EMA superinstance does not exceed the latency threshold.

[0032] Optionally, the EMAS management node sets the default latency specification to cost priority, that is, when the tenant does not select a latency specification, the default latency specification is cost priority. Since resources with low utilization rates are usually lower in price, the EMAS management node can schedule lower-priced resources for tenants based on the resource usage of the underlying physical server, thereby fully utilizing the idle resources of the data center.

[0033] Optionally, the placement strategy includes affinity and anti-affinity, wherein affinity means that the target EMA super instance and the first instance are located in the same resource node, and anti-affinity means that the target EMA super instance and the first instance are not located in the same resource node.

[0034] It should be noted that if the tenant selects a placement strategy, they must also enter the ID of the first instance so that the EMAS management node can determine the deployment location of the first instance. If the tenant does not enter a placement strategy, the EMAS management node can determine the resource node where the compute and memory resources of the target EMA super instance are located based on the resource usage of the underlying physical server.

[0035] Optionally, the access policy includes private, shared read-only, and shared read-write. Private means that when the number of target instances is 1, the target instance has read and write permissions for the target EMA superinstance. Shared read-only means that when the number of target instances mounted on the target EMA superinstance is greater than or equal to 2, the primary mounted instance among the at least two target instances has read and write permissions, while the other target instances have only read-only permissions. Shared read-write means that when the number of target instances mounted on the target EMA superinstance is greater than or equal to 2, both target instances have read and write permissions.

[0036] It should be noted that if the tenant only enters the ID of one target instance, the access policy is set to private by default. In addition, if the access policy is set to shared read-only, the tenant also needs to enter the ID of the primary mounting instance.

[0037] Optionally, usage modes include full peer mode, semi-peer mode and heterogeneous mode. Among them, full peer mode means that GuestOS fully manages, that is, the total memory of the second instance is composed of the local memory of the second instance and the memory of the target EMA super instance, and the local memory of the second instance and the memory of the target EMA super instance are uniformly addressed. Semi-peer mode means that GuestOS partially manages, that is, the total memory of the second instance is composed of the local memory of the second instance and the memory of the target EMA super instance, and the local memory of the second instance and the memory of the target EMA super instance are uniformly addressed, but the user can know the address segment corresponding to the local memory of the second instance and the address segment corresponding to the memory of the target EMA super instance. In heterogeneous mode, GuestOS does not manage the EMA super instance memory pool and does not uniformly address it.

[0038] In the second case, the target cloud service is provided in an instance-free manner. At this time, the performance configuration information of the target cloud service includes the service type, performance specifications, the maximum number of simultaneous connections and the rate cap. The service type indicates the tenant's business demand for memory resources and computing resources, the performance specifications indicate the number of target EMA super instances, the maximum number of simultaneous connections indicates the number of instances that the target EMA super instance can connect to at the same time, and the rate cap indicates the upper limit of the resources provided by the cloud platform to the tenant.

[0039] Optional service types include memory expansion, computing offloading, and caching. Different service types have different performance requirements for the target EMA superinstance. Caching requires high bandwidth but less demanding computing resources and latency specifications. Memory expansion requires both computing resources and memory capacity, while computing offloading has less demanding computing resources. Of course, in actual applications, other service types may be used to reflect specific business scenarios.

[0040] Optionally, the performance specifications include standard, high performance, and best effort. Under the standard specification, the number of target EMA superinstances is related to the number of instances connected to the target EMA superinstance, and the number of instances connected to the target EMA superinstance is greater than the number of target EMA superinstances. Under the high performance specification, the number of instances connected to the target EMA superinstance is equal to the number of EMA superinstances, i.e., each instance connected to the target EMA superinstance corresponds to one target EMA superinstance. Under the best effort specification, there is only one target EMA superinstance, i.e., all instances connected to the target EMA superinstances share one target EMA superinstance.

[0041] It should be noted that if the tenant does not fill in the above performance specifications and maximum number of simultaneous connections, the EMAS management node can dynamically adjust the number of target EMA super instances and the maximum number of simultaneous connections based on the resource usage of the underlying physical server.

[0042] Optionally, the unit of the rate cap may be RMB per hour, RMB per month, or RMB per year.

[0043] Optionally, after the tenant determines the above service type, performance specifications, and maximum number of simultaneous connections, the EMAS management node can determine the basic price corresponding to the service type, performance specifications, and maximum number of simultaneous connections configured by the tenant according to the relevant algorithm, and the upper limit of the rate filled in by the tenant is greater than or equal to the basic price.

[0044] Optionally, the EMAS management node can obtain the attribute information of the available resources of each resource node in each resource pool, and determine the target resource node from the multiple resource nodes based on the attribute information of the available resources of each resource node in each resource pool and the attribute information of the target cloud service. The resource nodes in the target resource node are located in the same resource pool, and an elastic memory acceleration instance EMAI is created in each resource node included in the target resource node to obtain one or more EMAIs, and a target EMA super instance is generated based on the one or more EMAIs.

[0045] Optionally, the configuration parameters of the target cloud service can be determined based on the attribute information of the target cloud service, and the target resource node can be determined from multiple resource nodes based on the configuration parameters of the target cloud service and the attribute information of the available resources of each resource node in each resource pool.

[0046] In the case of different target cloud service provisioning methods, the methods for determining the configuration parameters of the target cloud service are different, which will be introduced below.

[0047] In the first case, the target cloud service is provided in a super instance provision mode. In this case, the performance configuration information of the target cloud service can be directly used as the configuration parameters of the target cloud service.

[0048] In the second case, the target cloud service is provisioned without instances. In this case, the EMAS management node stores the mapping between the service type and the target cloud service's configuration parameters, including guaranteed capacity, burst capacity, computing resources, capacity step size, bandwidth, latency specification, guarantee mode, and placement strategy. Based on the mapping between the service type and the target cloud service's configuration parameters, the EMAS management node can determine the target cloud service's configuration parameters based on the tenant's configured service type.

[0049] It should be noted that in addition to obtaining the attribute information of the target cloud service from the configuration interface, the attribute information of the above target cloud service can also be obtained through API calls, email interactions, work order interactions, etc.

[0050] In actual applications, after providing the management interface to the tenant, the EMAS management node can also determine whether the resources of the target EMA super-instance are overloaded based on the tenant's resource usage of the target EMA super-instance. When it is determined that the resources of the target EMA super-instance are overloaded, based on the attribute information of the target cloud service, it is determined whether the resources provided by the cloud platform for the target EMA super-instance have reached the configured resource upper limit; when the resources provided by the cloud platform for the target EMA super-instance do not reach the configured resource upper limit, the target resource node is adjusted.

[0051] Optionally, the resources of the target EMA superinstance include memory resources and computing resources. Based on the tenant's usage of the target EMA superinstance's memory resources, it is determined whether the target EMA superinstance's memory resources are overloaded. Based on the computing resource usage of the target EMA superinstance, it is determined whether the computing resources of the target EMA superinstance are overloaded. If it is determined that the memory resources and / or computing resources of the target EMA superinstance are overloaded, it is determined that the resources of the target EMA superinstance are overloaded.

[0052] Optionally, the memory resource usage includes a memory occupancy ratio. If the memory occupancy ratio is greater than a memory ratio threshold, it is determined that the memory resource of the target EMA superinstance is overloaded; otherwise, it is determined that the memory resource of the target EMA superinstance is not overloaded.

[0053] Optionally, the computing resource usage includes a computing resource occupancy ratio, a number of queued requests, and a response time. If the computing resource occupancy ratio is less than a computing resource ratio threshold, the number of queued requests is less than a queue number threshold, and the response time is less than a response time threshold, then it is determined that the computing resources of the target EMA superinstance are not overloaded; otherwise, it is determined that the computing resources of the target EMA superinstance are overloaded.

[0054] Of course, in actual applications, the computing resource usage may also include at least one of the computing resource occupancy ratio, the number of queued requests, and the response time.

[0055] Optionally, when the resources provided by the cloud platform for the target EMA super instance reach the configured resource upper limit, a prompt message is sent to the tenant, where the prompt message is used to indicate that the resources provided by the cloud platform for the target EMA super instance have reached the configured resource upper limit.

[0056] In a second aspect, a cloud platform is provided, wherein the cloud platform has the function of implementing the method and behavior of providing cloud services in the first aspect. The cloud platform includes at least one module, wherein the at least one module is used to implement the method and behavior of providing cloud services in the first aspect.

[0057] In a third aspect, a computing device cluster is provided, wherein the computing device cluster includes at least one computing device, each computing device includes a processor and a memory, and the processor of the at least one computing device is used to execute instructions stored in the memory of the at least one computing device, so that the computing device cluster performs the method for providing cloud services described in the first aspect.

[0058] Optionally, each computing device may further include a communication bus, which is used to establish a connection between the processor and the memory of each computing device.

[0059] In a fourth aspect, a computer-readable storage medium is provided, wherein the storage medium includes computer program instructions. When the computer program instructions are executed by a computing device cluster, the computing device cluster executes the steps of the method for providing cloud services described in the first aspect.

[0060] In a fifth aspect, a computer program product comprising instructions is provided. When the instructions are executed by a computing device cluster, the computing device cluster executes the steps of the method for providing cloud services described in the first aspect.

[0061] The technical effects obtained in the above-mentioned second, third, fourth and fifth aspects are similar to those obtained by the corresponding technical means in the first aspect, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] FIG1 is a schematic diagram of an EMA super-instance provided in an embodiment of the present application;

[0063] FIG2 is a schematic diagram of an implementation environment provided by an embodiment of the present application;

[0064] FIG3 is a schematic diagram of another implementation environment provided by an embodiment of the present application;

[0065] FIG4 is a schematic structural diagram of a POD resource management node provided in an embodiment of the present application;

[0066] FIG5 is a flow chart of a method for providing cloud services provided in an embodiment of the present application;

[0067] FIG6 is a schematic diagram of resource distribution of an EMA super instance provided in an embodiment of the present application;

[0068] FIG7 is a schematic diagram of an EMA super instance mounting method provided in an embodiment of the present application;

[0069] FIG8 is a schematic diagram of another EMA super instance mounting method provided in an embodiment of the present application;

[0070] FIG9 is a schematic diagram of a non-instance providing method provided in an embodiment of the present application;

[0071] FIG10 is a schematic diagram of another non-example providing method provided in an embodiment of the present application;

[0072] FIG11 is a flowchart of using a target cloud service provided by an embodiment of the present application;

[0073] FIG12 is a schematic diagram of a non-instance providing method provided in an embodiment of the present application;

[0074] FIG13 is a flowchart of a Java application performance optimization service provided in an embodiment of the present application;

[0075] FIG14 is a schematic diagram of a Java application performance optimization service provided in an embodiment of the present application;

[0076] FIG15 is a schematic diagram of a data caching service provided in an embodiment of the present application;

[0077] FIG16 is a schematic structural diagram of an EMAS management node provided in an embodiment of the present application;

[0078] FIG17 is an architecture diagram of a cloud platform provided in an embodiment of the present application;

[0079] FIG18 is a schematic structural diagram of a computer device provided in an embodiment of the present application;

[0080] FIG19 is an architectural diagram of a computing device cluster provided in an embodiment of the present application;

[0081] FIG20 is a schematic diagram of a connection between computing devices provided in an embodiment of the present application. DETAILED DESCRIPTION

[0082] In order to make the objectives, technical solutions and advantages of this application clearer, the implementation methods of this application will be further described in detail below with reference to the accompanying drawings.

[0083] For ease of understanding, before explaining in detail the method for providing cloud services provided in the embodiments of the present application, the nouns, application scenarios and implementation environments involved in the embodiments of the present application are first introduced.

[0084] Virtual machine / virtual server (virtual machine): also known as a cloud server in a cloud scenario, is a complete computer system with complete hardware system functions running in a completely isolated environment on the same physical server / host based on virtualization technology and simulated by software.

[0085] Elastic cloud server: A virtual machine or cloud server managed by the elastic computing service in cloud computing. It has the characteristics of instant expansion and instant release. It can adjust computing resources at any time according to business needs and achieve horizontal expansion and contraction. It is also called an ECS instance.

[0086] A super instance is a virtualized logical instance. It is exposed as a single logical instance, but its underlying topology may consist of one or more homogeneous or heterogeneous instances. This underlying topology is transparent to the user, who can only directly operate the super instance itself, providing a user experience similar to that of a single instance.

[0087] It should be noted that when a superinstance consists of only one instance, the superinstance is equivalent to the instance. That is, when there is only one instance at the bottom layer of the superinstance, the user can operate on the instance.

[0088] Instanceless: This refers to a type of virtual resource provisioning method. When virtual resources are provided through instanceless, the underlying implementation relies on superinstances. That is, the virtual resource consists of one or more homogeneous or heterogeneous instances. The underlying topology is transparent to the user, and the superinstance is also transparent to the user. Users cannot directly operate the underlying instances or superinstances.

[0089] Elastic Memory Accelerator Super Instance (EMASI): A super instance managed by the Elastic Memory Accelerator Service (EMAS) management node. An EMASI is a logical instance with memory resources. Its memory capacity is composed of the memory resources of at least one Elastic Memory Accelerator Instance (EMAI). For ease of description, EMASIs are referred to as EMA super instances in the following text.

[0090] For example, referring to Figure 1, the EMA super-instance has 128 GB of memory resources. This EMA super-instance's memory resources are provided by EMAI1_1 on host01, EMAI1_2 on host02, and EMAI1_3 and EMAI1_4 on host03. EMAI1_1 provides 16 GB of memory resources, EMAI1_2 provides 16 GB of memory resources, EMAI1_3 provides 32 GB of memory resources, and EMAI1_4 provides 16 GB of memory resources.

[0091] EMAI: Instances that make up the EMA superinstance. EMAI is a physical instance and cannot be deployed across physical servers.

[0092] Guest operating system (Guest OS): refers to the operating system running in the virtual cloud server.

[0093] Host node operating system (Node OS): The operating system of the host node where the virtual machine or container is located.

[0094] A software development kit (SDK) is a software package that integrates a set of development tools to help developers develop applications for a specific platform. SDKs typically include tools such as compilers, debuggers, libraries, API documentation, and sample code. These tools help developers develop applications more efficiently and ensure that applications interact well with the platform at runtime. For example, a specific operating system or development language may have a corresponding SDK to help developers develop applications for that platform.

[0095] With the continuous development of information technology, especially the advancement of the internet, large government, enterprises, and financial institutions have begun to build their own data centers to meet their business resource needs. As data centers rapidly expand in size, their complexity and energy consumption are increasing year by year. Simply adding physical equipment is extremely expensive. Against this backdrop, cloud computing emerged. Leveraging virtualization technology, cloud service providers integrate resources and provide unified services, significantly improving resource utilization efficiency.

[0096] In the related art, cloud service providers provide tenants with basic resource services at the Infrastructure as a Service (Iaas) layer through virtual machines, containers, etc. Taking virtual machines as an example, the specifications of virtual machines are predefined patterns, and tenants need to specify the size of the computing resources and memory resources of the virtual machine from the preset specifications, that is, the number of central processing units (CPUs) and the size of the memory. If the specification of the virtual machine specified by the tenant from the preset specifications is 4U8G, this specification means that the virtual machine has 4 CPU cores and 8 gigabytes of memory, and the resources of the tenant cannot exceed the specified specifications when using the virtual machine. Compared with virtual machines, containers are more flexible. After selecting the number of CPUs for the container, the tenant can configure the memory capacity according to a certain step size, but the step size is fixed, and the ratio of the number of CPUs to the memory capacity also needs to be within the preset ratio range. For example, the preset ratio range is between 1:4 and 1:8. In this case, if the tenant selects 4 CPU cores as computing resources, the memory capacity selection range can only be 4GB-32GB, and the step size is 1GB. In other words, the user can only select an integer memory capacity and cannot select a memory capacity of 4.5GB.

[0097] However, in cloud computing scenarios, with the development of business and the evolution of applications, tenants have an increasing demand for memory capacity. The current preset specifications often cannot meet the needs of some tenants. If memory resources are insufficient, the computing performance of cloud services will be seriously affected.

[0098] Furthermore, the memory resources currently provided by the IaaS layer are essentially homogeneous, with identical memory performance specifications and address space performance specifications. Identical memory performance specifications mean that, given the same computing resources, virtual machines with different memory specifications have identical memory performance. Identical address space performance specifications mean that, for virtual machines with the same specifications, the read and write speeds for each memory address in the virtual machine are also consistent. Therefore, given the same computing resources, the performance of memory resources with different memory capacities remains unchanged, and memory resources with larger memory capacities command higher prices. If tenants purchase larger memory capacities as their memory needs increase, this significantly increases their costs. Furthermore, if the business itself does not require high memory performance and the tenant only needs to expand memory capacity, they may be unwilling to pay the additional cost for high-performance memory capacity. Furthermore, in real life, some businesses do not require high computing resource performance but have high requirements for memory capacity and performance. In this case, if tenants purchase both computing resources and memory resources at the same time, the utilization rate of the computing resources purchased by the tenants will be very low, and the memory capacity will also be very limited. As a result, the tenants will have to pay for unused computing resources and their memory capacity needs will not be met.

[0099] If a cloud service provider provides tenants with a preset specification with a larger memory ratio, it will seriously affect the efficiency of computing resource utilization in the physical server. For example, if a tenant is provided with a 1:6 virtual machine in a 1:4 physical server (the ratio of computing resources to memory resources is 1:4), the utilization rate of the physical server's computing resources will be reduced, resulting in resource waste and increased operation and maintenance costs for the cloud service provider. For example, physical server A has 100 CPU cores and 400GB of memory, that is, the specification of this physical server is 1:4. If a 1:6 virtual machine is issued to physical server A, then physical server A will have 66 CPU cores and 396GB of memory available to serve tenants. Since computing resources cannot usually be sold individually, in this case, 34 CPU cores and 4GB of memory in physical server A will not be able to serve tenants, resulting in resource waste and increasing operation and maintenance costs for the cloud service provider. At the same time, there are still some idle memory resources in the cloud service provider's data center that have not been sold yet. The overall memory resource sales rate is not high, but most of these unsold memory resources are distributed relatively scattered, that is, distributed on different physical servers, and cannot form a complete memory block to be sold together with the virtual machine or container, resulting in insufficient resource utilization and waste.

[0100] Based on this, an embodiment of the present application provides a method for providing cloud services. The EMAS management node can provide a configuration interface and obtain attribute information of the target cloud service from the configuration interface. The attribute information includes the provision method of the target cloud service and the performance configuration information of the target cloud service. Based on the performance configuration information, the EMAS management node can create a target EMA super instance in the target resource node, and then set the management interface of the target EMA super instance based on the provision method of the target cloud service, and provide the management interface to the tenant to provide the target cloud service to the tenant through the target EMA super instance. Since the attribute information is input by the tenant according to business needs, that is, the tenant does not need to be restricted by the preset specifications, and can directly customize the computing resources and memory resources of the target cloud service according to its own business needs, thereby meeting the needs of the tenant and improving the performance of the cloud service. In addition, since the target resource node includes at least one resource node, that is, the embodiment of the present application can integrate the resources of at least one resource node to provide services to the tenant. In this way, while meeting user needs, it can also effectively improve the utilization rate of idle resources in the data center, reduce resource waste and the operation and maintenance costs of the cloud service provider.

[0101] Please refer to Figure 2, which is a schematic diagram of an implementation environment provided by an embodiment of the present application. The implementation environment includes a tenant side 10 and a service side (cloud platform) 20. The service side 20 includes an EMAS management node 21 (not shown in Figure 2) and multiple resource nodes 22 (not shown in Figure 2). Optionally, the EMAS management node 21 includes a first control node 211, a data persistence layer 212, a billing system 213, and an application deployment channel 214.

[0102] Among them, the first control node 211 provides a configuration interface 2111 and displays it to the tenant side. The configuration interface 2111 is used to prompt the tenant to input the attribute information of the target cloud service to be provided according to business needs, and then the first control node 211 can obtain the attribute information of the target cloud service, and based on the attribute information of the target cloud service, create a target EMA super instance in the target resource node included in the multiple resource nodes 22 and set the management interface of the target EMA super instance, and provide the management interface to the tenant. The first control node 211 can also store the relevant data of the target cloud service, that is, convert the relevant data of the target cloud service into a data model that can be stored in the data persistence layer 212, and store it in the data persistence layer 212 to achieve long-term storage and access to the data, thereby completing the storage of the data. The data persistence layer 212 can be a disk or a database, and this embodiment of the application does not limit this. After the target EMA super instance is created, the first control node 211 can connect the target EMA super instance to the billing system 213 to realize billing for the target EMA super instance. After obtaining the management interface of the target EMA super instance, the tenant can deploy the required application through the management interface of the target EMA super instance and the application deployment channel 214 .

[0103] Please refer to Figure 3, which is a schematic diagram of another implementation environment provided by an embodiment of the present application, which includes an EMAS management node 21 (not shown in Figure 3) and multiple resource nodes 22 (the multiple resource nodes 22 are schematically represented by resource node 1 and resource node 2 in Figure 3). In addition to including a first control node 211, a data persistence layer 212, a billing system 213, and an application deployment channel 214, the EMAS management node 21 also includes at least one POD resource management node 215 (schematically illustrated as one POD resource management node in Figure 3), a cluster resource management node 216, a hot demounting management node 217, and a DC (Data Center) topology management node 218.

[0104] The multiple resource nodes 22 are located in at least one resource pool, a resource pool can be called a POD, each POD corresponds to a POD resource management node 215, each POD includes at least one resource node, the resource nodes in the same POD are located in the same ultra-high-speed network plane, each resource node can be a physical server, or it can be a combination of part of the resources (computing and memory) in the physical server.

[0105] For any resource node in any POD, the resource node includes a Node OS, an EMAS virtualization layer, and an EMAS agent. The EMAS virtualization layer is used to provide the virtualization interface (also known as the Application Programming Interface (API)) of the resource node and manage the resources in the resource node. The EMAS Agent is usually deployed on the physical server where the resource node is located, and is used to report the resource status of the resource node to the POD resource management node 214 to which the resource node belongs. The EMAS Agent is also used to create an EMA super instance by calling the API provided by the EMAS virtualization layer when the resource node is determined to be the target resource node.

[0106] The POD resource management node 215 manages resources within the POD, maintains the topology of resources within the POD, receives resource status reports from the EMAS Agent, and integrates the resources of the resource nodes within the POD resource management node 215 into a resource pool. The POD resource management node 215 can also report the resource status of the resource pool within the POD to the cluster resource management node and instruct the target resource node to create an EMAI.

[0107] In some embodiments, the resource pool may include a computing resource pool and a memory resource pool, each resource node may provide memory resources and / or computing resources, a resource node capable of providing memory resources is referred to as a memory resource node, and a resource node capable of providing computing resources is referred to as a computing resource node. The memory resources of each memory resource node may be accumulated into a memory resource pool. Computing resources may be isolated according to computing resource nodes, and the computing resources of different computing resource nodes are not accumulated and pooled to obtain computing resource pools corresponding to a plurality of computing resource nodes. Of course, in practical applications, the computing resources of the computing resource nodes may also be accumulated into a computing resource pool, and this embodiment of the application does not limit this. It should be noted that the above-mentioned ultra-high-speed network plane may be built by technologies such as Compute Express Link (CXL) and Remote Direct Memory Access (RDMA), and this embodiment of the application does not limit this.

[0108] In some embodiments, the first control node 211 can receive CRUD (Create, Retrieve, Update, and Delete), mount, and detach requests of the EMA super instance, and manage the life cycle of the EMA super instance based on the request, that is, perform operations such as creation, read query, update, delete, mount, and detach on the EMA super instance. The first control node 211 can also set a management interface for the EMA super instance and provide the management interface to the tenant. In some embodiments, the first resource node 211 can send a resource application or resource release request to the cluster resource management node 216. The resource application request includes the performance configuration information of the target cloud service, and the resource release request includes the identifier of the EMA super instance to be released. The cluster resource management node 216 can determine the target resource node based on the performance configuration information of the target cloud service and the resource status of the resource pool in each POD, and send the scheduling result (identifier of the target resource node) to the first control node 211. The cluster resource management node 216 can also determine the resource node that needs to release resources based on the identifier of the EMA superinstance, and send the scheduling result (the identifier of the resource node that needs to release resources) to the first control node 211. The first control node 211 receives the scheduling result sent by the cluster resource management node 216 and, based on the scheduling result, sends instance creation, update, deletion, and other requests to the POD resource management node 215. Taking the example of the first control node 211 sending a superinstance creation request to the POD resource management node 215, the instance creation request includes the identifier of the target resource node. Based on the identifier of the target resource node, the POD resource management node 215 sends an instruction to create an EMAI to the target resource node.

[0109] Optionally, the first control node 211 can also perform metadata management, quotas, and authentication. Metadata management refers to the inventory, integration, and management of metadata such as technical metadata and operational metadata. Quotas limit the number of cloud products (services) a tenant can purchase, preventing tenants from purchasing a large number of a single cloud product and causing insufficient underlying resources. Authentication verifies the user's identity, ensuring that only authorized or privileged users can access cloud services or use cloud resources.

[0110] The hot detach management node 217 manages the hot detachment of EMA super-instances. Specifically, when an EMA super-instance is attached to instance A, it detaches the EMA super-instance without impacting the services of instance A. The DC topology management node 218 manages and optimizes the physical topology of the data center to improve its reliability and efficiency.

[0111] It should be noted that the above-mentioned cloud platform can be a server cluster or distributed system composed of multiple physical servers, or it can be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), as well as big data and artificial intelligence platforms, or it can be a cloud computing service center.

[0112] The above-mentioned first control node 211, data persistence layer 212, billing system 213, application deployment channel 214, POD resource management node 215, cluster resource management node 216, thermal demounting management node 217, and DC topology management node 218 can be the same node, or they can be different nodes, or at least two of the above-mentioned multiple nodes can constitute the same node. This embodiment of the present application does not limit this.

[0113] Those skilled in the art should understand that the above-mentioned EMAS management nodes, multiple resource nodes and cloud platforms are only examples. Other existing or future EMAS management nodes, multiple resource nodes and cloud platforms that are applicable to the embodiments of the present application should also be included in the scope of protection of the embodiments of the present application and are included here by reference.

[0114] It should be noted that the application scenarios and implementation environments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Ordinary technicians in this field can know that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0115] Please refer to Figure 4, which is a structural diagram of a POD resource management node provided in an embodiment of the present application. The POD resource management node includes a POD resource allocator, a POD memory global allocator, a POD computing power pool, a POD resource global monitor and a POD resource global scheduler.

[0116] Among them, the POD resource allocator receives a super instance creation request sent by the first control node 211. The instance creation instruction includes a target resource node identifier and occupancy information. The occupancy information indicates the resource size of each target resource node that the current super instance needs to occupy, and sends the target resource node identifier and occupancy information to the POD memory global addresser and the POD computing power pool respectively. The POD memory global addresser records the memory resource occupancy based on the occupancy information, and the POD computing power pool records the occupancy of computing resources based on the occupancy information. The POD resource allocator can also send an instruction to create an EMAI to the virtualization layer of the target resource node. The POD memory global allocator can also integrate the memory resources of the resource nodes within the POD, and record the memory model, rate, bandwidth and other information of the resource nodes, and then perform unified addressing to realize the registration of the resource nodes, and perform hierarchical management of the resource nodes through data structures such as bitmaps and hash tables. The POD computing power pool can also integrate the computing resources of the resource nodes within the POD, and record the CPU architecture, generation, main frequency, number of cores and other information of the resource nodes to realize the registration of the resource nodes. The POD resource global monitor is to monitor the CPU, memory utilization, and bandwidth occupancy of the resource nodes within the POD to identify resource nodes with high resource occupancy (overload), i.e., hotspots. The POD resource global scheduler can reallocate resources across resource nodes based on the hotspots identified by the POD resource global monitor and the real-time global resource status sent by the POD memory global addresser and the POD computing power pool.

[0117] Next, the method for providing cloud services provided in the embodiment of the present application is explained in detail.

[0118] FIG5 is a flowchart of a method for providing cloud services according to an embodiment of the present application. The method is applied to an EMAS management node included in a cloud platform, which also includes multiple resource nodes configured to provide at least one of computing resources and memory resources. Referring to FIG5 , the method includes the following steps.

[0119] Step 501: Provide a configuration interface, which is used to prompt the tenant to input attribute information of the target cloud service to be provided according to business requirements. The attribute information of the target cloud service includes the provision method of the target cloud service and the performance configuration information of the target cloud service.

[0120] In some embodiments, the EMAS management node prompts the tenant to input the attribute information of the target cloud service to be provided according to business needs by providing a configuration interface. The tenant inputs the provision method and performance configuration information of the target cloud service based on the configuration interface.

[0121] As an example, the target cloud service may be provided in a super instance provision mode or a non-instance provision mode. The tenant may select either of the two as the target cloud service provision mode.

[0122] The super-instance provisioning method means that the target EMA super-instance used to provide the target cloud service to the tenant is visible to the tenant, and the target EMA's lifecycle is managed by the tenant. The non-instance provisioning method means that the target EMA super-instance is invisible to the tenant, and the target EMA's lifecycle is not managed by the tenant.

[0123] When the target cloud service selected by the tenant is provided in different ways, the performance configuration information of the target cloud service includes different contents, which will be introduced below.

[0124] In the first case, the target cloud service is provided as a super-instance. In this case, the target cloud service's performance configuration information includes usage type, memory capacity, capacity guarantee mode, capacity step, computing resources, and bandwidth. Usage type indicates whether the target EMA super-instance provides services in a mounted form; memory capacity indicates the target EMA super-instance's memory capacity; capacity guarantee mode indicates whether the target EMA super-instance's memory resources are distributed across variable resource nodes; capacity step indicates the memory size of each EMAI included in the target EMA super-instance; computing resources indicates the target EMA super-instance's computing performance; and bandwidth indicates the target EMA super-instance's data transfer volume per unit time.

[0125] In some embodiments, the usage type includes single instance usage and mounted usage. Single instance usage means that the target EMA super instance does not provide services in a mounted form. Mounted usage means that the target EMA super instance provides services in a mounted form.

[0126] It should be noted that the tenant can directly select one of the two parameters, single instance use or mounted use, to determine the usage type. In actual applications, the tenant can also determine the usage type of the target cloud service by selecting the corresponding image for single instance use or mounted use. In other words, the usage type of the target cloud service is determined based on the image type. This embodiment of the application does not limit this.

[0127] Optionally, when the usage type is single instance, the tenant can use the target EMA super instance as a special cloud server, namely, a low-cost, extra-large memory instance. The tenant can mount data volumes, attach virtual network cards, or add elastic IP addresses to the target EMA super instance.

[0128] In some embodiments, the memory capacity is divided into guaranteed capacity and burst capacity, wherein the guaranteed capacity refers to the minimum memory capacity required to be guaranteed by the target EMA super instance, and the burst capacity indicates the maximum capacity of the target EMA super instance, which is greater than or equal to the guaranteed capacity.

[0129] It should be noted that, when billing, the larger the guaranteed capacity, the higher the price; and when the guaranteed capacity is the same, the greater the difference between the guaranteed capacity and the burst capacity, the higher the price.

[0130] In some embodiments, capacity guarantee modes include dynamic and static modes. Dynamic mode is a rotational mode, meaning that the EMAS management node can split and relocate the EMAIs included in the target EMA super-instance based on the resource usage of the underlying physical server. Static mode is a non-rotational mode, meaning that after the EMASI is created, it cannot be split or relocated.

[0131] Splitting means using at least two EMAIs to provide resources originally provided by a single EMAI. For example, if EMAI1 on resource node A provides 8GB of memory resources, after splitting, EMAI12 on resource node B can provide 3GB of memory resources, and EMAI3 on ​​resource node C can provide 5GB of memory resources. In other words, the 8GB of memory resources are ultimately provided by two EMAIs.

[0132] Relocation means that resources originally provided by the first resource node are provided by other resource nodes besides the first resource node. For example, if EMASI1 on resource node A provides 8GB of memory resources, after relocation, EMAI2 on resource node B can provide 8GB of memory resources.

[0133] Optionally, the default capacity guarantee mode of the EMAS management node is dynamic mode, that is, when the tenant does not select the capacity guarantee mode, the default capacity guarantee mode is dynamic mode, so that resources can be scheduled according to the resource usage of the underlying physical server, thereby achieving full utilization of resources.

[0134] It should be noted that when billing, the price of dynamic mode is higher than that of static mode.

[0135] In some embodiments, the capacity step indicates the minimum memory capacity of each EMAI included in the target EMA superinstance. That is, the memory capacity of each EMAI is no less than the capacity step, and the capacity step sets a lower limit on the memory capacity specification of all EMAIs comprising the EMA superinstance. In this case, the number of EMAIs comprising the EMA superinstance is no greater than the value obtained by dividing the actual memory capacity of the EMA superinstance by the capacity step.

[0136] For example, the target EMA superinstance requires 128G memory capacity, and the capacity step is set to 32G. Then the memory capacity of the EMAIs that constitute the target EMA superinstance is greater than or equal to 32G. In this case, the target EMA superinstance consists of a maximum of 4 EMAIs.

[0137] Optionally, if a tenant sets the capacity step size to 0, this does not mean that the memory resources corresponding to each EMASI are 0. Instead, it means that the EMAS management node can create EMASIs for the tenant based on the resource usage of the underlying physical servers, without being constrained by the capacity step size. The size of the EMASI memory resources is related to the actual resource size allocated by the EMAS management node. For example, the EMAS management node can prioritize creating EMASIs on resource nodes with smaller memory capacities based on the resource usage of the underlying physical servers.

[0138] It should be noted that the default capacity step of the EMAS management node is 0, that is, when the tenant does not select a capacity step, the EMAS management node can create an EMASI for the tenant based on the resource usage of the underlying physical server.

[0139] It should be noted that when billing, the larger the capacity step, the higher the price.

[0140] For example, bandwidth may refer to the amount of data transmitted per second by the target EMA super instance, in GB per second. Of course, in practical applications, bandwidth may also be represented in other forms, which is not limited in the embodiments of the present application.

[0141] It should be noted that when billing, the greater the bandwidth, the higher the price.

[0142] In some embodiments, the unit of computing resources is a virtual computing unit (VCU). One VCU can be a core in a physical CPU or a data processing unit (DPU), or it can be a hyperthread of a core in a physical CPU or DPU. This embodiment of the present application does not limit this.

[0143] It should be noted that the one VCU can be a computing resource used solely by the target EMA super instance, or it can be a computing resource shared with other EMA super instances, and this embodiment of the application does not limit this. For example, if the target EMA super instance includes the computing resources of 1VCU, EMA super instance 1 and the target EMA super instance can share the computing resources of the 1VCU. For example, referring to Figure 6, the memory resources of EMA super instance 01 are 128GB and the computing resources are 16VCU, the memory resources of EMA super instance 02 are 256GB and the computing resources are 16VCU, the usage type of EMA super instance 01 and EMA super instance 02 is single instance use, and EMA super instance 01 and EMA super instance 02 share the computing resources of 16VCU, and the memory resources of EMA super instance 01 and EMA super instance 02 are not shared.

[0144] Optionally, the default computing resource of the EMAS management node is 1 VCU, that is, when the tenant does not set the computing resource, the default computing resource is 1 VCU.

[0145] Based on the description above, usage types include single-instance usage and mounted usage. Depending on the usage type, the target cloud service's performance configuration information may also include different information, which will be explained below.

[0146] In case 1, the usage type is single instance usage. That is, the usage type indicates that the target EMA super instance does not provide services in a mounted form. In this case, the performance configuration information of the target cloud service also includes a latency specification and a placement strategy. The latency specification indicates the access latency of the target EMA super instance's memory resources, and the placement strategy indicates whether the target EMA super instance's memory resources and computing resources are located on the same resource node.

[0147] In some embodiments, the latency specification includes performance priority or cost priority. Performance priority means that the EMAS management node selects the resource node with the shortest latency from among multiple resource nodes as the target resource node. Cost priority means that the EMAS management node selects the resource node with the lowest price from among multiple resource nodes as the target resource node.

[0148] For example, if one target resource node is currently needed, but there are five resource nodes that all meet the tenant's configured memory capacity, capacity step size, computing resources, and bandwidth requirements, in this case, if the tenant chooses performance priority, the EMAS management node can select the resource node with the lowest memory resource access latency from the five resource nodes as the target resource node. If the tenant chooses cost priority, the EMAS management node can determine the price of each resource node according to the relevant algorithm and then select the resource node with the lowest price from the five resource nodes as the target resource node.

[0149] In practice, in addition to directly selecting performance or cost priority, tenants can also enter a specific latency threshold to determine the latency specification. In this case, the EMAS management node determines the target resource node based on the latency threshold, and the access latency of the memory resource of the target EMA super-instance is ultimately determined to be no greater than the latency threshold.

[0150] For example, the latency threshold entered by the tenant may be 100 milliseconds. If one target resource node is currently required, but there are five resource nodes that meet the tenant's configured memory capacity, capacity step, computing resources, and bandwidth requirements, in this case, a resource node with a latency less than or equal to the latency threshold can be selected from the five resource nodes as the target resource node.

[0151] It should be noted that if there are at least two resource nodes among the above-mentioned five resource nodes whose delays are less than or equal to the delay threshold, then a resource node can be randomly selected from the at least two resource nodes as the target resource node, or, the resource node with the lowest price can be selected from the two resource nodes as the target resource node, or, the resource node with the best performance can be selected from the two resource nodes as the target resource node. This embodiment of the present application does not limit this.

[0152] Optionally, the EMAS management node sets the default latency specification to cost priority, that is, when the tenant does not select a latency specification, the default latency specification is cost priority. Since resources with low utilization rates are usually lower in price, the EMAS management node can schedule lower-priced resources for tenants based on the resource usage of the underlying physical server, thereby fully utilizing the idle resources of the data center.

[0153] In some embodiments, the placement strategy includes affinity and anti-affinity, where affinity means that the memory resources and computing resources of the target EMA superinstance are located on the same resource node, and anti-affinity means that the memory resources and computing resources of the target EMA superinstance are not located on the same resource node.

[0154] It should be noted that if the tenant does not fill in the placement strategy, the EMAS management node can determine the resource node where the computing resources and memory resources of the target EMA super instance are located based on the resource usage of the underlying physical server.

[0155] Case 2: The usage type is mounted usage, that is, the usage type indicates that the target EMA super-instance provides services in a mounted form. In this case, the performance configuration information of the target cloud service also includes the identifier, latency specification, placement strategy, access strategy, and usage mode of at least one target instance. The target instance refers to the instance used to mount the target EMA super-instance. The latency specification indicates the access latency of the target instance to the target EMA super-instance. The placement strategy indicates whether the target EMA super-instance and a first instance are located on the same resource node. The first instance is one of the at least one target instance. The access strategy indicates the read and write permissions of the at least one target instance to the target EMA super-instance. The usage mode indicates the memory addressing method of the target EMA super-instance and a second instance is one of the at least one target instance.

[0156] It should be noted that the target instance can be a virtual machine instance, a bare metal server instance, or a container instance. The target instance needs to be created using an image that supports EMA capabilities.

[0157] In some embodiments, the latency specification includes performance priority or cost priority. Performance priority means that the EMAS management node selects the resource node with the shortest access latency to the target instance from multiple resource nodes as the target resource node. Cost priority means that the EMAS management node selects the resource node with the lowest cost from multiple resource nodes as the target resource node.

[0158] For example, if one target resource node is currently needed, but there are five resource nodes that all meet the tenant's configured memory capacity, capacity step, computing resources, and bandwidth requirements, in this case, if the tenant chooses performance priority, the EMAS management node can use the relevant algorithm to select the resource node with the lowest access latency to the target instance from the five resource nodes as the target resource node. If the tenant chooses cost priority, the EMAS management node can use the relevant algorithm to determine the price of each resource node and then select the resource node with the lowest price from the five resource nodes as the target resource node.

[0159] In practice, in addition to directly selecting performance or cost priority, tenants can also enter a specific latency threshold to determine the latency specification. In this case, the EMAS management node determines the target resource node based on the latency threshold, ensuring that the target instance's access latency to the target EMA superinstance does not exceed the latency threshold.

[0160] For example, the latency threshold entered by the tenant may be 100 milliseconds. If one target resource node is currently required, but there are five resource nodes that meet the tenant's configured memory capacity, capacity step, computing resources, and bandwidth requirements, in this case, a resource node whose access latency to the target instance is less than or equal to the latency threshold can be selected from the five resource nodes as the target resource node.

[0161] It should be noted that if there are at least two resource nodes among the above-mentioned five resource nodes whose delays are less than or equal to the delay threshold, then a resource node can be randomly selected from the at least two resource nodes as the target resource node, or, the resource node with the lowest price can be selected from the two resource nodes as the target resource node, or, the resource node with the best performance can be selected from the two resource nodes as the target resource node. This embodiment of the present application does not limit this.

[0162] Optionally, the EMAS management node sets the default latency specification to cost priority, that is, when the tenant does not select a latency specification, the default latency specification is cost priority. Since resources with low utilization rates are usually lower in price, the EMAS management node can schedule lower-priced resources for tenants based on the resource usage of the underlying physical server, thereby fully utilizing the idle resources of the data center.

[0163] In some embodiments, the placement strategy includes affinity and anti-affinity, wherein affinity means that the target EMA superinstance and the first instance are located on the same resource node, and anti-affinity means that the target EMA superinstance and the first instance are not located on the same resource node.

[0164] It should be noted that if the tenant selects a placement strategy, they must also enter the ID of the first instance so that the EMAS management node can determine the deployment location of the first instance. If the tenant does not enter a placement strategy, the EMAS management node can determine the resource node where the compute and memory resources of the target EMA super instance are located based on the resource usage of the underlying physical server.

[0165] In some embodiments, access policies include private, shared read-only, and shared read-write. Private means that when the number of target instances is 1, the target instance has read and write permissions on the target EMA superinstance. Shared read-only means that when the number of target instances mounted on the target EMA superinstance is greater than or equal to 2, the primary mounted instance of the at least two target instances has read and write permissions, while the other target instances have only read-only permissions. Shared read-write means that when the number of target instances mounted on the target EMA superinstance is greater than or equal to 2, both target instances have read and write permissions.

[0166] It should be noted that if the tenant only enters the ID of one target instance, the access policy is set to private by default. In addition, if the access policy is set to shared read-only, the tenant also needs to enter the ID of the primary mounting instance.

[0167] In some embodiments, usage modes include full peer mode, semi-peer mode and heterogeneous mode. Among them, full peer mode means that the GuestOS is fully managed, that is, the total memory of the second instance is composed of the local memory of the second instance and the memory of the target EMA super instance, and the local memory of the second instance and the memory of the target EMA super instance are uniformly addressed. Semi-peer mode means that the GuestOS is partially managed, that is, the total memory of the second instance is composed of the local memory of the second instance and the memory of the target EMA super instance, and the local memory of the second instance and the memory of the target EMA super instance are uniformly addressed, but the user can know the address segment corresponding to the local memory of the second instance and the address segment corresponding to the memory of the target EMA super instance. In heterogeneous mode, the GuestOS does not manage the EMA super instance memory pool and does not uniformly address it. The EMA super instance is similar to another non-uniform memory access architecture (NUMA) node, and the tenant or authorized user is responsible for resource management.

[0168] Optionally, an EMA superinstance can be mounted to a virtual machine or container. Two EMA superinstances can be mounted to the same virtual machine or container. For example, referring to Figure 7 , the virtual machine or container is mounted with a 2-terabyte (TB) Elastic Volume Service (EVS), namely a 2-TB cloud hard drive, EMA superinstance 1, and EMA superinstance 2. In some embodiments, EMA superinstance 3 can also support sharing. Referring to Figure 8 , EMA superinstance 3 can be mounted to two virtual machines or containers simultaneously.

[0169] In the second case, the target cloud service is provided in an instance-free manner. At this time, the performance configuration information of the target cloud service includes the service type, performance specifications, the maximum number of simultaneous connections and the rate cap. The service type indicates the tenant's business demand for memory resources and computing resources, the performance specifications indicate the number of target EMA super instances, the maximum number of simultaneous connections indicates the number of instances that the target EMA super instance can connect to at the same time, and the rate cap indicates the upper limit of the resources provided by the cloud platform to the tenant.

[0170] In some embodiments, service types include memory expansion, computing power offloading, and caching. Different service types have different performance requirements for the target EMA super instance. Among them, caching requires high bandwidth and low requirements for computing resources and latency specifications. Memory expansion has requirements for both computing resources and memory capacity, and computing power offloading has low requirements for computing resources. Of course, in actual applications, there may be other service types that can reflect business scenarios, and the embodiments of this application do not limit this.

[0171] For example, performance specifications include standard, high performance, and best effort. Under the standard specification, the number of target EMA superinstances is related to the number of instances connected to the target EMA superinstance, and the number of instances connected to the target EMA superinstance is greater than the number of target EMA superinstances. Under the high performance specification, the number of instances connected to the target EMA superinstance is equal to the number of EMA superinstances, i.e., each instance connected to the target EMA superinstance corresponds to one target EMA superinstance. Under the best effort specification, there is only one target EMA superinstance, i.e., all instances connected to the target EMA superinstances share one target EMA superinstance.

[0172] It should be noted that if the tenant does not fill in the above performance specifications and maximum number of simultaneous connections, the EMAS management node can dynamically adjust the number of target EMA super instances and the maximum number of simultaneous connections based on the resource usage of the underlying physical server.

[0173] Optionally, the unit of the rate cap can be RMB per hour, RMB per month, or RMB per year. The embodiment of the present application does not limit the billing period of the rate cap.

[0174] In some embodiments, after the tenant determines the above-mentioned service type, performance specifications, and maximum number of simultaneous connections, the EMAS management node can determine the basic price corresponding to the service type, performance specifications, and maximum number of simultaneous connections configured by the tenant according to the relevant algorithm, and the upper limit of the rate filled in by the tenant is greater than or equal to the basic price.

[0175] Optionally, the EMAS management node stores the correspondence between the service type and the service price, the correspondence between the performance specification and the first price coefficient, and the correspondence between the maximum number of simultaneous links and the second price coefficient. Based on the service type configured by the tenant, the target service price corresponding to the service type configured by the tenant can be determined. Based on the performance specification and the maximum number of simultaneous links configured by the tenant, the target first price coefficient and the target second price coefficient can be determined from the correspondence between the performance specification and the first price coefficient, and the correspondence between the maximum number of simultaneous links and the second price coefficient. The target service price is multiplied by the target first price coefficient and then by the target second price coefficient to obtain the base price corresponding to the service type, performance specification, and maximum number of simultaneous links configured by the tenant. Of course, the above method of determining the base price is only an example. In actual applications, the base price can also be determined by other methods, and the embodiments of the present application do not limit this.

[0176] Step 502: Obtain the attribute information of the target cloud service from the configuration interface, and based on the attribute information of the target cloud service, create a target EMA super instance in the target resource node included in the multiple resource nodes. The target EMA super instance is used to provide the target cloud service for the tenant. The target resource node includes at least one resource node, and the available resources of the target resource node meet the business needs of the tenant.

[0177] Obtain attribute information of available resources of each resource node in each resource pool, determine a target resource node from the multiple resource nodes based on the attribute information of available resources of each resource node in each resource pool and the attribute information of the target cloud service, each resource node in the target resource node is located in the same resource pool, create an elastic memory acceleration instance EMAI in each resource node included in the target resource node to obtain one or more EMAIs, and generate a target EMA super instance based on the one or more EMAIs.

[0178] In some embodiments, the configuration parameters of the target cloud service can be determined based on the attribute information of the target cloud service, and the target resource node can be determined from multiple resource nodes based on the configuration parameters of the target cloud service and the attribute information of the available resources of each resource node in each resource pool.

[0179] In the case of different target cloud service provisioning methods, the methods for determining the configuration parameters of the target cloud service are different, which will be introduced below.

[0180] In the first case, the target cloud service is provided in a super instance provision mode. In this case, the performance configuration information of the target cloud service can be directly used as the configuration parameters of the target cloud service.

[0181] In the second case, the target cloud service is provisioned without instances. In this case, the EMAS management node stores the mapping between the service type and the target cloud service's configuration parameters, including guaranteed capacity, burst capacity, computing resources, capacity step size, bandwidth, latency specification, guarantee mode, and placement strategy. Based on the mapping between the service type and the target cloud service's configuration parameters, the EMAS management node can determine the target cloud service's configuration parameters based on the tenant's configured service type.

[0182] Based on the attribute information of the available resources of each resource node in each resource pool and the configuration parameters of the target cloud service, there are many ways to determine the target resource node from multiple resource nodes. In some embodiments, first fit, best fit and other algorithms can be used to determine the target resource node, which is not limited in the embodiments of the present application.

[0183] It should be noted that in addition to obtaining the attribute information of the target cloud service from the configuration interface, the attribute information of the above-mentioned target cloud service can also be obtained through API calls, email interactions, work order interactions, etc., and this embodiment of the present application does not limit this.

[0184] Step 503: Based on the provision mode of the target cloud service, a management interface of the target EMA super-instance is set, and the management interface of the target EMA super-instance is provided to the tenant.

[0185] In some embodiments, the EMAS management node can send an SDK corresponding to the target cloud service provision method to the tenant, and the tenant can use the target EMA super instance based on the SDK, and the SDK can indicate how to use the target EMA super instance.

[0186] In actual applications, if the target cloud service is provisioned using the instanceless provisioning method, the EMAS management node can create a target instanceless service node and return the identifier of the target instanceless service node to the tenant. The underlying layer of this target instanceless service node is the target EMA superinstance. In this case, the SDK corresponding to the instanceless provisioning method includes the EMA server installation package. Tenants can install the EMA server for instances that need to connect to the target EMA superinstance and, based on the SDK and the identifier of the target instanceless service node, configure the EMA server's backend target to the target instanceless service node.

[0187] For example, referring to FIG9 , when the target cloud service is provided in an instanceless provisioning mode, the resources of the target instanceless service node can come from the resources of the memory resource node and the computing resource node. In some embodiments, the resources of the target instanceless service node can also be resources collected from the DPU or other devices by the idle resource collection node, such as the idle resources of the acceleration node and the network node. Referring to FIG10 , the target instanceless service node still provides resources to the target EMA super instance, but the target EMA super instance itself is transparent to the tenant. The EMAS management node can gradually allocate resources to the target EMA super instance. That is, when there is a request and the rate cap is met, the target EMA super instance is allocated resources according to the resources required by the request. If there is no request, no resources are allocated to the target EMA super instance. That is, the resources allocated by the EMAS management node to the target EMA super instance depend on the actual resource amount required by the tenant. As long as it is within the rate cap, the EMAS management node can continuously and gradually provide resources. Once the rate cap is reached, the request is rejected or the historical data in the memory is replaced with new data through memory page swapping technology (swap) and the least recently used (LRU) algorithm.

[0188] Next, a method for providing cloud services provided by an embodiment of the present application will be described again through three examples.

[0189] For the first example, please refer to Figure 11. The tenant first creates four ECS instances, namely virtual machine 01, virtual machine 02, virtual machine 03, and virtual machine 04, through the ECS management node included in the cloud platform. The images of these four ECS instances support EMA capabilities. Then, the tenant can enter the attribute information of the target cloud service in the configuration interface provided by the EMAS management node. Among them, the service provision mode of the target cloud service is instance-free provision mode. The performance configuration information of the target cloud service includes the service type of computing power offloading, the performance specification of standard, the maximum number of simultaneous connections of 4, and the rate cap of 20.98 yuan / hour. Please refer to Figure 12. The EMAS management node creates a target instanceless service node 01 and an EMA super-instance 01 based on the target cloud service's attribute information, and returns the target instanceless service node's identifier 01 and the SDK corresponding to the instanceless provisioning method to the tenant. Based on the SDK corresponding to the instanceless provisioning method and the target instanceless service node's identifier 01, the tenant can install the EMA server on the ECS instance and set the EMA server's backend target to the target instanceless service node, thereby connecting the four ECS instances to the target instanceless service node. This allows the tenant to offload computing operations that need to be offloaded to the EMA super-instance 01. These computing operations can include network packet processing, encryption and decryption, and other input / output operations. It should be noted that Figure 12 is merely an example. In actual applications, the backend of the target instanceless service node 01 may not necessarily be an EMA super-instance. The number of EMA super-instances can be flexibly adjusted based on the performance configuration information of the target cloud service set by the tenant.

[0190] In the second example, if a tenant needs to use ECS instance 1 to execute the computing work brought by the Java application deployed on the ECS instance 1, the ECS instance 1 itself also needs to perform the memory garbage collection function, and the execution of the garbage collection function will affect the efficiency of business execution. In order to improve the performance of Java applications, the cloud platform can provide a Java Runtime Environment Accelerating Service (JAS) based on a virtual machine, also known as a Java application performance optimization service. The JAS can provide services to tenants in combination with a method for providing cloud services provided in an embodiment of the present application.

[0191] In this case, the cloud platform includes JAS management nodes, ECS management nodes, and EMAS management nodes, please refer to Figure 13. The tenant purchases ECS instance 1 through the JAS management node and enters the uninstall service type, which refers to the type of computing that ECS instance 1 needs to uninstall. In the above scenario, the uninstall service type is the garbage collection uninstall service. The JAS management node can call the ECS management node to create ECS instance 1. The ECS management node uses an image that contains a Java Runtime Environment (JRE) and supports JAS capabilities to create ECS instance 1 and ECS instance 2. Among them, ECS instance 1 is used to execute the computing work brought by the Java application deployed on the ECS instance 1, and ECS instance 2 is used to perform garbage collection functions.

[0192] In some embodiments, the JAS management node stores a correspondence between the offloading service type and the attribute information of the target cloud service, which includes the provision method of the target cloud service and the target cloud service performance configuration information. The performance configuration information includes guaranteed capacity, burst capacity, capacity guarantee mode, bandwidth, latency specification, access policy and usage type. The JAS management node can determine the attribute information of the target cloud service from the correspondence. Among them, the provision method of the target cloud service corresponding to the garbage collection offloading service is the super instance provision method, the guaranteed capacity is 64GB, the burst capacity is 64GB, the capacity guarantee mode is static mode, the bandwidth is 12GB per second, the latency specification is performance priority, the access policy is shared read and write, the usage type is mounted use, the target instances are ECS instance 1 and ECS instance 2, and the placement strategy is affinity with ECS instance 2. The JAS management node can determine the attribute information of the target cloud service based on the garbage collection unloading service, and then send the attribute information of the target cloud service to the EMAS management node. Please refer to Figure 14. The EMAS management node can create an EMA super instance 1 based on the attribute information of the target cloud service. The EMA super instance 1 and the ECS instance 2 are located in the same resource node, namely resource node 1. The JAS management node calls the EMAS management node to mount the EMA super instance 1 to ECS instance 1 and ECS instance 2.

[0193] Tenants can directly operate ECS instance 1, which is responsible for the application's computing work. The local memory of ECS instance 1 acts as a high-speed memory buffer. EMA super instance 1 stores complete local memory information for ECS instance 1. ECS instance 2 can perform garbage collection on the memory in EMA super instance 1 and synchronize the garbage collection results to the local memory of ECS instance 1, enabling ECS ​​instance 2 to assist ECS instance 1 in completing garbage collection.

[0194] Considering the insufficient computing resource performance of EMA superinstance 1, in the second example, ECS instance 2 is created and garbage collection is performed on ECS instance 2. EMA superinstance 1 is used to synchronize the local memory information of ECS instance 1. In actual applications, if the computing resource performance of EMA superinstance 1 is higher, there is no need to create ECS instance 2, and garbage collection can be performed directly on EMA superinstance 1.

[0195] For the third example, please refer to Figure 15. The cloud platform can provide a data caching service through which tenants can purchase a highly cost-effective memory caching service. The data caching service can provide services to tenants in combination with a method for providing cloud services provided in an embodiment of the present application. The data caching service can be a memory architecture and system (memArts), and in actual applications it can also be other types of data caching services, which are not limited in this embodiment of the present application.

[0196] Tenants purchase memory cache services through the data cache service management node and specify the properties of the target cloud service. The target cloud service is provided as a super instance, with a guaranteed capacity of 128GB, a burst capacity of 128GB, a dynamic capacity guarantee mode, a bandwidth of 16GB per second, a cost-first latency specification, a private access policy, and a mounted usage type. In this case, the data cache service management node calls the EMAS management node to create EMA super instance 2. Because the latency specification is cost-first, the target resource node prioritizes the resource node with the smallest resource specification among the resource nodes that meet the properties of the target cloud service. The data cache service control plane calls the ECS management node to create an ECS instance for the tenant as a cache controller. The cache controller instance is responsible for maintaining the synchronization of data in the cache with data in the persistence layer. The cache controller instance uses the image that supports the data cache service provided by the data cache service to ensure that the ECS contains the server of the data cache service. The data cache service management node calls the EMAS management node to mount the EMA super instance to the cache controller instance. The EMAS management node sets the management interface of the target EMA super instance based on the provision method of the target cloud service, and sends the SDK corresponding to the provision method of the target cloud service to the tenant. The tenant can use the target EMA super instance based on the SDK.

[0197] In actual applications, after providing the management interface to the tenant, the EMAS management node can also determine whether the resources of the target EMA super-instance are overloaded based on the tenant's resource usage of the target EMA super-instance. When it is determined that the resources of the target EMA super-instance are overloaded, based on the attribute information of the target cloud service, it is determined whether the resources provided by the cloud platform for the target EMA super-instance have reached the configured resource upper limit; when the resources provided by the cloud platform for the target EMA super-instance do not reach the configured resource upper limit, the target resource node is adjusted.

[0198] In some embodiments, the resources of the target EMA superinstance include memory resources and computing resources. Based on the tenant's usage of the target EMA superinstance's memory resources, it is determined whether the target EMA superinstance's memory resources are overloaded. Based on the usage of the computing resources of the target EMA superinstance, it is determined whether the computing resources of the target EMA superinstance are overloaded. If it is determined that the memory resources and / or computing resources of the target EMA superinstance are overloaded, it is determined that the resources of the target EMA superinstance are overloaded.

[0199] In some embodiments, the memory resource usage includes a memory occupancy ratio. If the memory occupancy ratio is greater than a memory ratio threshold, it is determined that the memory resources of the target EMA superinstance are overloaded; otherwise, it is determined that the memory resources of the target EMA superinstance are not overloaded.

[0200] The memory ratio threshold is pre-set and ranges from 0 to 1. For example, the memory ratio threshold can be set to 0.8. This means that if the target EMA superinstance's memory resource usage ratio exceeds 0.8, the target EMA superinstance is considered memory overloaded. In practice, the memory ratio threshold can be adjusted based on actual needs.

[0201] In some embodiments, the computing resource usage includes a computing resource occupancy ratio, a number of queued requests, and a response time. If the computing resource occupancy ratio is less than a computing resource ratio threshold, the number of queued requests is less than a queue number threshold, and the response time is less than a response time threshold, then it is determined that the computing resources of the target EMA superinstance are not overloaded; otherwise, it is determined that the computing resources of the target EMA superinstance are overloaded.

[0202] Of course, in actual applications, the computing resource usage may also include at least one of the computing resource occupancy ratio, the number of request queues, and the response time, which is not limited in the embodiments of the present application.

[0203] The computing resource ratio threshold, the queue quantity threshold, and the response time threshold are pre-set. In actual applications, the computing resource ratio threshold, the queue quantity threshold, and the response time threshold can also be adjusted according to needs.

[0204] Depending on the target cloud service provisioning method, the implementation methods for determining whether the resources provided by the cloud platform for the target EMA super instance have reached the configured resource limit may vary. These methods are described below.

[0205] When the target cloud service is provided in the non-instance provision mode, the resource configuration information provided by the current cloud platform for the target EMA super-instance is obtained. Based on the resource configuration information, the configuration price corresponding to the resource configuration information is determined according to a relevant algorithm. If the configuration price is greater than or equal to the rate cap, it is determined that the resources provided by the cloud platform for the target EMA super-instance have reached the configured resource cap. Otherwise, it is determined that the resources provided by the cloud platform for the target EMA super-instance have not reached the configured resource cap.

[0206] Since the resource overload of the target EMA super-instance means that the memory resources and / or computing resources of the target EMA super-instance are overloaded, when the target cloud service is provided in the super-instance provision mode, if the type of overloaded resources is different, the method of determining whether the resources provided by the cloud platform for the target EMA super-instance reaches the configured resource upper limit is different, which will be introduced below.

[0207] When it is determined that the memory resources of the target EMA super instance are overloaded, the resource configuration information provided by the current cloud platform for the target EMA super instance is obtained. The configuration information includes the deployed memory capacity. If the deployed memory capacity is greater than or equal to the burst memory capacity, it is determined that the resources provided by the cloud platform for the target EMA super instance have reached the configured resource upper limit.

[0208] When it is determined that the computing resources of the target EMA super instance are overloaded, the resource configuration information provided by the current cloud platform for the target EMA super instance is obtained. The configuration information includes the deployed computing resource specifications. If the deployed computing resource specifications are greater than or equal to the computing resource specifications configured by the tenant, it is determined that the resources provided by the cloud platform for the target EMA super instance have reached the configured resource upper limit.

[0209] When it is determined that the memory resources and computing resources of the target EMA super-instance are overloaded, obtain the resource configuration information provided by the current cloud platform for the target EMA super-instance, which includes the deployed computing resource specifications and deployed memory capacity. If the deployed computing resource specifications are greater than or equal to the computing resource specifications configured by the tenant and / or the deployed memory capacity is greater than or equal to the burst memory capacity, it is determined that the resources provided by the cloud platform for the target EMA super-instance have reached the configured resource upper limit.

[0210] Optionally, when the resources provided by the cloud platform for the target EMA super instance reach the configured resource upper limit, a prompt message is sent to the tenant, where the prompt message is used to indicate that the resources provided by the cloud platform for the target EMA super instance have reached the configured resource upper limit.

[0211] When the target cloud service is provided in a super-instance manner and the target EMA super-instance provides services in a mounted form, if the instance on which the target EMA super-instance is mounted also has other EMA super-instances or system volumes (for example, EVS) mounted, and the remaining storage capacity of the other EMA super-instances or system volumes is greater than or equal to the amount of data in the target super-instance, the tenant can also trigger a hot unmount instruction. In response to the hot unmount instruction, the EMAS management node can migrate the data in the target EMA super-instance to the migration destination, thereby realizing the hot unmounting of the target EMA super-instance. The migration destination is other EMA super-instances or system volumes mounted on the instance on which the target EMA super-instance is mounted except for the target EMA super-instance.

[0212] In some embodiments, the implementation process of migrating the data in the target EMA super instance to the migration destination includes: setting the status of the target EMA super instance to being unhooked, packetizing the data in the target EMA super instance according to a relevant algorithm to obtain multiple data packets, marking the data attributes corresponding to the multiple data packets, the data attributes indicating the frequency of access of the data in the data packets, and migrating the multiple data packets to the migration destination based on the data attributes of the multiple data packets.

[0213] The data attributes include cold data, warm data, and hot data. The frequency of cold data access is lower than the frequency of warm data access, and the frequency of warm data access is lower than the frequency of hot data access.

[0214] Based on the data attributes of the multiple data packets, the implementation process of migrating the multiple data packets to the migration destination includes: packaging and copying the data packets with data attributes of cold data and warm data to the migration destination, and copying the data packets with data attributes of hot data to the migration destination in multiple times according to relevant algorithms.

[0215] It should be noted that if the instance on which the target EMA super-instance is mounted has other EMA super-instances and a system volume mounted on it, the other EMA super-instance will be prioritized as the migration destination.

[0216] In the embodiment of the present application, since the attribute information of the target cloud service is input by the tenant according to business needs, that is, the tenant does not need to be restricted by preset specifications and can directly customize the computing resources and memory resources of the target cloud service according to its own business needs, thereby meeting the needs of the tenant and improving the performance of the cloud service. In addition, since the target resource node includes multiple resource nodes, that is, the embodiment of the present application can integrate the resources of at least one resource node to provide services to the tenant. In this way, while meeting user needs, it can also effectively improve the utilization rate of idle resources in the data center, reduce resource waste and reduce the operation and maintenance costs of the cloud service provider.

[0217] Figure 16 is a schematic diagram of the structure of an EMAS management node provided in an embodiment of the present application. The EMAS management node is included in a cloud platform, which also includes multiple resource nodes. Referring to Figure 16 , the EMAS management node includes: a provision module 1601, an acquisition module 1602, a creation module 1603, and a configuration module 1604.

[0218] Providing module 1601 is used to provide a configuration interface. This configuration interface prompts the tenant to enter the attribute information of the target cloud service to be provided according to business requirements. This attribute information includes the provision method of the target cloud service and the performance configuration information of the target cloud service. The detailed implementation process is referred to the corresponding content of each embodiment above and is not repeated here.

[0219] The acquisition module 1602 is used to obtain the attribute information of the target cloud service from the configuration interface. The detailed implementation process is referred to the corresponding content of each embodiment above and will not be repeated here.

[0220] Creation module 1603 is configured to create a target elastic memory acceleration (EMA) super-instance in a target resource node included in the plurality of resource nodes based on the attribute information of the target cloud service. The target EMA super-instance is used to provide the target cloud service for the tenant. The target resource node includes at least one resource node, and the available resources of the target resource node meet the service requirements. The detailed implementation process is referred to the corresponding content in the above embodiments and is not repeated here.

[0221] The setting module 1604 is used to set the management interface of the target EMA super instance based on the provision mode of the target cloud service and provide the management interface to the tenant. The detailed implementation process is referred to the corresponding content of each embodiment above and will not be repeated here.

[0222] Optionally, the target cloud service is provided in a super-instance provision mode, which means that the target EMA super-instance is visible to the tenant and the life cycle of the target EMA is managed by the tenant; the performance configuration information of the target cloud service includes usage type, memory capacity, capacity guarantee mode, capacity step, computing resources and bandwidth; the usage type indicates whether the target EMA super-instance provides services in a mounted form, the memory capacity indicates the memory capacity of the target EMA super-instance, the capacity guarantee mode indicates whether the resource nodes to which the memory resources of the target EMA super-instance are distributed are variable, the capacity step indicates the memory size of each elastic memory acceleration instance EMAI included in the target EMA super-instance, the computing resources indicate the computing performance of the target EMA super-instance, and the bandwidth indicates the amount of data transmitted by the target EMA super-instance per unit time.

[0223] Optionally, when the usage type indicates that the target EMA super-instance does not provide services in a mounted form, the performance configuration information of the target cloud service also includes a latency specification and a placement strategy, wherein the latency specification indicates the access latency of the memory resources of the target EMA super-instance, and the placement strategy indicates whether the memory resources and computing resources of the target EMA super-instance are located in the same resource node.

[0224] Optionally, when the usage type indicates that the target EMA super-instance provides services in a mounted form, the performance configuration information of the target cloud service also includes the identifier, latency specification, placement strategy, access strategy and usage mode of at least one target instance, where the target instance refers to the instance used to mount the target EMA super-instance, the latency specification indicates the access latency of the target instance to the target EMA super-instance, the placement strategy indicates whether the target EMA super-instance and the first instance are located on the same resource node, the first instance is one of the at least one target instance, the access strategy indicates the read and write permissions of the at least one target instance to the target EMA super-instance, and the usage mode indicates the memory addressing method of the target EMA super-instance and the second instance, and the second instance is one of the at least one target instance.

[0225] Optionally, the target cloud service is provided in an instance-free provision mode, which means that the target EMA super-instance is invisible to the tenant and the life cycle of the target EMA is not managed by the tenant; the performance configuration information of the target cloud service includes the service type, performance specifications, the maximum number of simultaneous connections and the rate cap; the service type indicates the tenant's business demand for memory resources and computing resources, the performance specifications indicate the number of target EMA super-instances, the maximum number of simultaneous connections indicates the number of instances that the target EMA super-instance can connect to at the same time, and the rate cap indicates the upper limit of the resources provided by the cloud platform to the tenant.

[0226] Optionally, the multiple resource nodes are located in at least one resource pool; the creation module 1603 is specifically configured to:

[0227] Obtain the attribute information of available resources of each resource node in each resource pool;

[0228] Determine a target resource node from a plurality of resource nodes based on attribute information and performance configuration information of available resources of each resource node in each resource pool, wherein each resource node in the target resource node is located in the same resource pool;

[0229] Creating an elastic memory acceleration instance EMAI in each resource node included in the target resource node to obtain one or more EMAIs;

[0230] Generates a target EMA superinstance based on one or more EMAIs.

[0231] Optionally, the EMAS management node also includes:

[0232] A first determination module is configured to determine whether the resources of the target EMA super instance are overloaded based on the resource usage of the target EMA super instance by the tenant;

[0233] The second determination module is configured to determine, when it is determined that the resources of the target EMA super instance are overloaded, whether the resources provided by the cloud platform for the target EMA super instance have reached the configured resource upper limit based on the attribute information of the target cloud service;

[0234] The adjustment module is used to adjust the target resource node when the resources provided by the cloud platform for the target EMA super instance do not reach the configured resource upper limit.

[0235] Optionally, the EMAS management node also includes:

[0236] The prompt module is used to send a prompt message to the tenant when the resources provided by the cloud platform for the target EMA super instance reach the configured resource upper limit. The prompt message is used to indicate that the resources provided by the cloud platform for the target EMA super instance have reached the configured resource upper limit.

[0237] In the embodiment of the present application, since the attribute information of the target cloud service is input by the tenant according to business needs, that is, the tenant does not need to be restricted by preset specifications and can directly customize the computing resources and memory resources of the target cloud service according to its own business needs, thereby meeting the needs of the tenant and improving the performance of the cloud service. In addition, since the target resource node includes multiple resource nodes, that is, the embodiment of the present application can integrate the resources of at least one resource node to provide services to the tenant. In this way, while meeting user needs, it can also effectively improve the utilization rate of idle resources in the data center, reduce resource waste and reduce the operation and maintenance costs of the cloud service provider.

[0238] It should be noted that the EMAS management node device provided in the above embodiment is only used as an example to illustrate the division of the above functional modules when providing cloud services. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the node can be divided into different functional modules to complete all or part of the functions described above. In addition, the EMAS management node provided in the above embodiment and the method embodiment for providing cloud services are based on the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0239] The provisioning module, acquisition module, creation module, and configuration module in the EMAS management node shown in Figure 16 can all be implemented via software, hardware, or a combination of software and hardware. The following describes the implementation of the provisioning module in a device providing cloud services as an example. Similarly, the implementation of the other modules described above can refer to the implementation of the provisioning module.

[0240] As an example of a software functional unit, a module may include code running on a computing instance. The computing instance may include at least one of a physical host, a virtual machine, and a container. Furthermore, the computing instance may be one or more. For example, a module may include code running on multiple hosts / virtual machines / containers. It should be noted that the multiple hosts / virtual machines / containers used to run the code may be distributed in the same region or in different regions. Furthermore, the multiple hosts / virtual machines / containers used to run the code may be distributed in the same availability zone (AZ) or in different AZs, and each AZ includes one data center or multiple geographically close data centers. Generally, a region may include multiple AZs.

[0241] Similarly, multiple hosts / virtual machines / containers running the code can be distributed within the same VPC or across multiple virtual private clouds (VPCs). Typically, a VPC is set up within a region. Cross-region communication between two VPCs within the same region, or between VPCs in different regions, requires a communication gateway within each VPC to interconnect the VPCs.

[0242] As an example of a hardware functional unit, a module may include at least one computing device, such as a server. Alternatively, the module may be implemented using an application-specific integrated circuit (ASIC) or a programmable logic device (PLD). The PLD may be a complex programmable logical device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof.

[0243] The multiple computing devices included in the provisioning module can be distributed in the same region or in different regions. The multiple computing devices included in the provisioning module can be distributed in the same AZ or in different AZs. Similarly, the multiple computing devices included in the provisioning module can be distributed in the same VPC or in multiple VPCs. The multiple computing devices can be any combination of servers, ASICs, PLDs, CPLDs, FPGAs, GALs, and other computing devices.

[0244] In other embodiments, the providing module in the EMAS management node can be used to execute any step in the above-mentioned method of providing cloud services, the acquiring module in the EMAS management node can be used to execute any step in the above-mentioned method of providing cloud services, the creating module in the EMAS management node can be used to execute any step in the above-mentioned method of providing cloud services, and the setting module in the EMAS management node can be used to execute any step in the above-mentioned method of providing cloud services. The steps that the providing module, acquiring module, creating module and setting module in the EMAS management node are responsible for implementing can be specified as needed. Different steps in the above-mentioned method of providing cloud services are respectively implemented by the providing module, acquiring module, creating module and setting module in the EMAS management node, thereby realizing the full functions of providing the EMAS management node.

[0245] FIG17 is an architectural diagram of a cloud platform provided in an embodiment of the present application. The cloud platform includes an EMAS management node and multiple resource nodes, each of which is configured to provide at least one of computing resources and memory resources. The EMAS management node and the multiple resource nodes are connected in a communication manner.

[0246] The EMAS management node is used to provide a configuration interface and obtain attribute information of the target cloud service from the configuration interface. Based on the attribute information, a target elastic memory acceleration EMA super instance is created in the target resource node included in multiple resource nodes. Then, based on the provision method of the target cloud service, a management interface of the target EMA super instance is set and provided to the tenant.

[0247] The EMAS management node and multiple resource nodes in Figure 17 can be implemented in software or hardware. For example, the implementation of the EMAS management node will be described below using the EMAS management node as an example. Similarly, the implementation of the resource nodes can refer to the implementation of the EMAS management node.

[0248] The EMAS management node may include code running on a computing instance. The computing instance may include at least one of a physical host (computing device), a virtual machine, and a container. Furthermore, the computing instance may be one or more. For example, the EMAS management node may include code running on multiple hosts / virtual machines / containers. It should be noted that the multiple hosts / virtual machines / containers used to run the code may be distributed in the same region or in different regions. Furthermore, the multiple hosts / virtual machines / containers used to run the code may be distributed in the same AZ or in different AZs, and each AZ includes one data center or multiple data centers with close geographical locations. Generally, a region may include multiple AZs.

[0249] Similarly, multiple hosts / virtual machines / containers running the code can be distributed within the same VPC or across multiple VPCs. Typically, a VPC is set up within a region. Cross-region communication between two VPCs within the same region, or between VPCs in different regions, requires a communication gateway within each VPC to interconnect the VPCs.

[0250] The EMAS management node may include at least one computing device, such as a server. Alternatively, the EMAS management node may be implemented using an ASIC or a PLD. The PLD may be implemented using a CPLD, FPGA, GAL, or any combination thereof.

[0251] The multiple computing devices included in an EMAS management node can be distributed in the same region or in different regions. They can also be distributed in the same AZ or in different AZs. Similarly, they can be distributed in the same VPC or across multiple VPCs. These multiple computing devices can be any combination of servers, ASICs, PLDs, CPLDs, FPGAs, GALs, and other computing devices.

[0252] The present application also provides a computing device 100. As shown in FIG18 , computing device 100 includes a bus 102, a processor 104, a memory 106, and a communication interface 108. Processor 104, memory 106, and communication interface 108 communicate with each other via bus 102. Computing device 100 may be a server or a terminal device. It should be understood that this application does not limit the number of processors and memories in computing device 100.

[0253] Bus 102 may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, among others. Buses may be classified as address buses, data buses, control buses, and the like. For ease of illustration, FIG18 shows only one line, but this does not imply that there is only one bus or only one type of bus. Bus 102 may include a path for transmitting information between various components of computing device 100 (e.g., memory 106, processor 104, and communication interface 108).

[0254] Processor 104 may include any one or more of a central processing unit (CPU), a graphics processing unit (GPU), a microprocessor (MP), a digital signal processor (DSP), an integrated circuit, and the like. The integrated circuit may include an ASIC, a programmable logic device (PLD), or a combination thereof. Optionally, the PLD may be a CPLD, an FPGA, a GAL, or any combination thereof.

[0255] The memory 106 may include volatile memory, such as random access memory (RAM). The memory 106 may also include non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD) or solid state drive (SSD), or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory 106 exists independently and is connected to the processor 104 via the bus 102, or the memory 106 is integrated with the processor 104.

[0256] The memory 106 stores executable program codes, and the processor 104 executes the executable program codes to implement the aforementioned method for providing cloud services. In other words, the memory 106 stores instructions for executing the method for providing cloud services.

[0257] For example, the memory 106 stores executable code, and the processor 104 executes the executable code to implement the functions of providing a module, obtaining a module, creating a module, and setting a module as shown in Figure 16 above. In other words, the memory 106 stores instructions for executing the method for providing cloud services provided in the embodiments of the present application.

[0258] The communication interface 108 uses a transceiver module such as, but not limited to, a network interface card or a transceiver to enable communication between the computing device 100 and other devices or communication networks. Network interfaces include wired communication interfaces or wireless communication interfaces. Examples of wired communication interfaces include Ethernet interfaces. Ethernet interfaces can be optical interfaces, electrical interfaces, or a combination thereof. Wireless communication interfaces can include wireless local area network (WLAN) interfaces, cellular network communication interfaces, or a combination thereof.

[0259] Embodiments of the present application also provide a computing device cluster. The computing device cluster includes multiple computing devices. The computing devices can be servers, such as central servers, edge servers, or local servers in a local data center. In some embodiments, the computing devices can also be terminal devices such as desktop computers, laptop computers, or smartphones.

[0260] As shown in Figure 19, the computing device cluster includes multiple computing devices 100. The same instructions for executing the above-mentioned method for providing cloud services may be stored in the memories 106 of the multiple computing devices 100 in the computing device cluster.

[0261] In some possible implementations, the memories 106 of the multiple computing devices 100 in the computing device cluster may also each store a portion of the instructions for executing the aforementioned method for providing cloud services. In other words, the combination of the multiple computing devices 100 may collectively execute all the instructions for executing the aforementioned method for providing cloud services.

[0262] It should be noted that the memory 106 in different computing devices 100 in the computing device cluster may store different instructions for executing partial functions of the above-described method for providing cloud services. In other words, the instructions stored in the memory 106 in different computing devices 100 may implement the functions of some or all of the modules included in the apparatus for providing cloud services described below.

[0263] In some possible implementations, one or more computing devices in a computing device cluster may be connected via a network. The network may be a wide area network or a local area network, etc. FIG20 shows a possible implementation. As shown in FIG16 , two computing devices 100A and 100B are connected via a network. Specifically, the network is connected via a communication interface in each computing device. In this type of possible implementation, the memory 106 in the computing device 100A stores instructions for executing the functions of a portion of the modules included in the EMAS management node described above. At the same time, the memory 106 in the computing device 100B stores instructions for executing the functions of another portion of the modules included in the EMAS management node described above.

[0264] It should be understood that the functions of the computing device 100A shown in FIG20 may also be completed by multiple computing devices 100. Similarly, the functions of the computing device 100B may also be completed by multiple computing devices 100.

[0265] The present application also provides another computing device cluster. The connection relationship between the computing devices in this computing device cluster can be similar to the connection method of the computing device cluster described in Figures 19 and 20. However, the memory 106 of one or more computing devices 100 in this computing device cluster can store the same instructions for executing the method for providing cloud services.

[0266] In some possible implementations, the memory 106 of one or more computing devices 100 in the computing device cluster may also store partial instructions for executing the method for providing cloud services. In other words, the combination of one or more computing devices 100 can jointly execute the instructions for executing the method for providing cloud services.

[0267] The embodiment of the present application also provides a computer-readable storage medium. The computer-readable storage medium can be any available medium that can be stored by a computing device or a data storage device such as a data center that contains one or more available media. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive). The computer-readable storage medium includes instructions that instruct the computing devices in the computer device cluster to execute the method for providing cloud services provided by the embodiment of the present application.

[0268] Embodiments of the present application also provide a computer program product including instructions. The computer program product may be software or a program product including instructions that can be executed on a computing device or stored on any available medium. When the computer program product is executed on at least one computing device cluster, the computing device cluster executes a method for providing cloud services.

[0269] In the above embodiments, all or part of the embodiments can be implemented by software, hardware, firmware, or any combination thereof. When implemented using software, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) method. The computer-readable storage medium can be any available medium that can be accessed by a computer, or a data storage device such as a server or data center that includes one or more available media integrated therein. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, or a magnetic tape), an optical medium (e.g., a digital versatile disc (DVD)), or a semiconductor medium (e.g., a solid state disk (SSD)). It is worth noting that the computer-readable storage medium mentioned in the embodiments of the present application may be a non-volatile storage medium, in other words, a non-transient storage medium.

[0270] It should be understood that the "plurality" mentioned herein refers to two or more. In the description of the embodiments of the present application, unless otherwise specified, " / " means or, for example, A / B can mean A or B; "and / or" in this article is merely a description of the association relationship of associated objects, indicating that there can be three relationships, for example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, in order to facilitate a clear description of the technical solutions of the embodiments of the present application, in the embodiments of the present application, words such as "first" and "second" are used to distinguish between identical or similar items with substantially the same functions and effects. Those skilled in the art will understand that words such as "first" and "second" do not limit the quantity and execution order, and words such as "first" and "second" do not necessarily limit them to be different.

[0271] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, stored data, displayed data, etc.), and signals involved in the embodiments of this application are all 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. For example, the attribute information of the target cloud service involved in the embodiments of this application is obtained with full authorization.

[0272] The above description is an embodiment provided for this application and is not intended to limit this application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of this application should be included in the scope of protection of this application.

Claims

1. A method for providing cloud services, characterized in that: Applied to an elastic memory acceleration service EMAS management node included in a cloud platform, the cloud platform also includes a plurality of resource nodes, the resource nodes are used to provide at least one of computing resources and memory resources, the method includes: Providing a configuration interface, wherein the configuration interface is used to prompt the tenant to input attribute information of the target cloud service to be provided according to business requirements, wherein the attribute information includes the provision method of the target cloud service and the performance configuration information of the target cloud service; Acquire attribute information of the target cloud service from the configuration interface; Based on the attribute information, create a target elastic memory acceleration EMA super instance in a target resource node included in the multiple resource nodes, the target EMA super instance is used to provide the target cloud service for the tenant, the target resource node includes at least one resource node, and the available resources of the target resource node meet the business requirements; Based on the provision mode of the target cloud service, a management interface of the target EMA super instance is set, and the management interface is provided to the tenant.

2. The method according to claim 1, characterized in that The target cloud service is provided in a super instance provision mode, which means that the target EMA super instance is visible to the tenant and the life cycle of the target EMA is managed by the tenant; The performance configuration information of the target cloud service includes usage type, memory capacity, capacity guarantee mode, capacity step, computing resources and bandwidth; the usage type indicates whether the target EMA super instance provides services in a mounted form, the memory capacity indicates the memory capacity of the target EMA super instance, the capacity guarantee mode indicates whether the resource nodes to which the memory resources of the target EMA super instance are distributed are variable, the capacity step indicates the memory size of each elastic memory acceleration instance EMAI included in the target EMA super instance, the computing resources indicate the computing performance of the target EMA super instance, and the bandwidth indicates the amount of data transmitted by the target EMA super instance per unit time.

3. The method according to claim 2, characterized in that When the usage type indicates that the target EMA super instance does not provide services in a mounted form, the performance configuration information of the target cloud service also includes a latency specification and a placement strategy, wherein the latency specification indicates the access latency of the memory resources of the target EMA super instance, and the placement strategy indicates whether the memory resources and computing resources of the target EMA super instance are located on the same resource node.

4. The method according to claim 2, characterized in that In the case where the usage type indicates that the target EMA super instance provides services in a mounted form, the performance configuration information of the target cloud service also includes an identifier, a latency specification, a placement strategy, an access strategy and a usage mode of at least one target instance, wherein the target instance refers to an instance used to mount the target EMA super instance, the latency specification indicates the access latency of the target instance to the target EMA super instance, the placement strategy indicates whether the target EMA super instance and the first instance are located on the same resource node, the first instance is one of the at least one target instance, the access policy indicates the read and write permissions of the at least one target instance to the target EMA super instance, and the usage mode indicates the memory addressing method of the target EMA super instance and the second instance, and the second instance is one of the at least one target instance.

5. The method according to claim 1, characterized in that The target cloud service is provided in an instance-free provision mode, which means that the target EMA super instance is invisible to the tenant and the life cycle of the target EMA is not managed by the tenant; The performance configuration information of the target cloud service includes service type, performance specification, maximum number of simultaneous connections and rate cap; The service type indicates the tenant's business demand for memory resources and computing resources, the performance specification indicates the number of target EMA super instances, the maximum number of simultaneous connections indicates the number of instances that the target EMA super instance can connect to at the same time, and the rate cap indicates the upper limit of resources provided by the cloud platform to the tenant.

6. The method according to claim 1, characterized in that The multiple resource nodes are located in at least one resource pool; and based on the performance configuration information, creating a target elastic memory acceleration EMA super instance in a target resource node included in the multiple resource nodes includes: Obtain the attribute information of available resources of each resource node in each resource pool; Based on the attribute information of the available resources of each resource node in each resource pool and the performance configuration information, Determine the target resource node in the point, each resource node in the target resource node is located in the same resource pool; Creating an elastic memory acceleration instance EMAI in each resource node included in the target resource node to obtain one or more EMAIs; Based on the one or more EMAIs, the target EMA superinstance is generated.

7. The method according to claim 1, characterized in that After setting the management interface of the target EMA super instance based on the provision mode of the target cloud service and providing the management interface to the tenant, the method further includes: Determining whether resources of the target EMA superinstance are overloaded according to resource usage of the target EMA superinstance by the tenant; In the case where it is determined that the resources of the target EMA super instance are overloaded, determining whether the resources provided by the cloud platform for the target EMA super instance have reached a configured resource upper limit based on the attribute information of the target cloud service; When the resources provided by the cloud platform for the target EMA super instance do not reach the upper limit of configured resources, the target resource node is adjusted.

8. The method according to claim 7, characterized in that The method further comprises: When the resources provided by the cloud platform for the target EMA super instance reach the upper limit of configured resources, a prompt message is sent to the tenant, where the prompt message is used to indicate that the resources provided by the cloud platform for the target EMA super instance have reached the upper limit of configured resources.

9. A cloud platform, characterized in that: The cloud platform includes an elastic memory acceleration service EMAS management node and a plurality of resource nodes, wherein the resource nodes are used to provide at least one of computing resources and memory resources; The EMAS management node is used to provide a configuration interface, wherein the configuration interface is used to prompt the tenant to input attribute information of the target cloud service to be provided according to business requirements, wherein the attribute information includes a provision method of the target cloud service and performance configuration information of the target cloud service; and obtain the attribute information of the target cloud service from the configuration interface; Based on the attribute information, create a target elastic memory acceleration EMA super instance in a target resource node included in the multiple resource nodes, the target EMA super instance is used to provide the target cloud service for the tenant, the target resource node includes at least one resource node, and the available resources of the target resource node meet the business requirements; Based on the provision mode of the target cloud service, a management interface of the target EMA super instance is set, and the management interface is provided to the tenant.

10. The cloud platform according to claim 9, characterized in that: The target cloud service is provided in a super instance provision mode, which means that the target EMA super instance is visible to the tenant and the life cycle of the target EMA is managed by the tenant; The performance configuration information of the target cloud service includes usage type, memory capacity, capacity guarantee mode, capacity step, computing resources and bandwidth; the usage type indicates whether the target EMA super instance provides services in a mounted form, the memory capacity indicates the memory capacity of the target EMA super instance, the capacity guarantee mode indicates whether the resource nodes to which the memory resources of the target EMA super instance are distributed are variable, the capacity step indicates the memory size of each elastic memory acceleration instance EMAI included in the target EMA super instance, the computing resources indicate the computing performance of the target EMA super instance, and the bandwidth indicates the amount of data transmitted by the target EMA super instance per unit time.

11. The cloud platform according to claim 10, characterized in that: When the usage type indicates that the target EMA super instance does not provide services in a mounted form, the performance configuration information of the target cloud service also includes a latency specification and a placement strategy, wherein the latency specification indicates the access latency of the memory resources of the target EMA super instance, and the placement strategy indicates whether the memory resources and computing resources of the target EMA super instance are located on the same resource node.

12. The cloud platform according to claim 10, characterized in that: In the case where the usage type indicates that the target EMA super instance provides services in a mounted form, the performance configuration information of the target cloud service also includes an identifier, a latency specification, a placement strategy, an access strategy and a usage mode of at least one target instance, wherein the target instance refers to an instance used to mount the target EMA super instance, the latency specification indicates the access latency of the target instance to the target EMA super instance, the placement strategy indicates whether the target EMA super instance and the first instance are located on the same resource node, the first instance is one of the at least one target instance, the access policy indicates the read and write permissions of the at least one target instance to the target EMA super instance, and the usage mode indicates the memory addressing method of the target EMA super instance and the second instance, and the second instance is one of the at least one target instance.

13. The cloud platform according to claim 9, characterized in that: The target cloud service is provided in an instance-free provision mode, which means that the target EMA super instance is invisible to the tenant and the life cycle of the target EMA is not managed by the tenant; The performance configuration information of the target cloud service includes service type, performance specification, maximum number of simultaneous connections and rate cap; The service type indicates the tenant's business demand for memory resources and computing resources, the performance specification indicates the number of target EMA super instances, the maximum number of simultaneous connections indicates the number of instances that the target EMA super instance can connect to at the same time, and the rate cap indicates the upper limit of resources provided by the cloud platform to the tenant.

14. The cloud platform according to claim 9, characterized in that: The plurality of resource nodes are located in at least one resource pool; The EMAS management node is used to obtain the attribute information of the available resources of each resource node in each resource pool; based on the attribute information of the available resources of each resource node in each resource pool and the performance configuration information, determine the target resource node from the multiple resource nodes, and each resource node in the target resource node is located in the same resource pool; create an elastic memory acceleration instance EMAI in each resource node included in the target resource node to obtain one or more EMAIs; based on the one or more EMAIs, generate the target EMA super instance.

15. The cloud platform according to claim 9, characterized in that: The EMAS management node is used to determine whether the resources of the target EMA super-instance are overloaded based on the resource usage of the target EMA super-instance by the tenant; when it is determined that the resources of the target EMA super-instance are overloaded, based on the attribute information of the target cloud service, determine whether the resources provided by the cloud platform for the target EMA super-instance have reached the upper limit of configured resources; when the resources provided by the cloud platform for the target EMA super-instance have not reached the upper limit of configured resources, adjust the target resource node.

16. The cloud platform according to claim 15, characterized in that: The EMAS management node is used to send a prompt message to the tenant when the resources provided by the cloud platform for the target EMA super instance reach the configured resource upper limit, and the prompt message is used to indicate that the resources provided by the cloud platform for the target EMA super instance have reached the configured resource upper limit.

17. A computing device cluster, characterized in that: The computing device cluster includes at least one computing device, each computing device including a processor and a memory; The processor of the at least one computing device is used to execute instructions stored in the memory of the at least one computing device, so that the computing device cluster executes the method according to any one of claims 1-8.

18. A computer-readable storage medium, characterized in that: The storage medium includes computer program instructions. When the computer program instructions are executed by a computing device cluster, the computing device cluster executes the method according to any one of claims 1 to 8.

19. A computer program product comprising instructions, characterized in that When the instructions are executed by a computing device cluster, the computing device cluster executes the method according to any one of claims 1 to 8.

Citation Information

Patent Citations

  • Method for providing cloud service, cloud platform and related device

    CN119071291A

  • Method for creating multi-tenant elastic file system based on cloud computing

    CN112035403A

  • Method, device and equipment for providing cloud service

    CN115131046A

  • Cloud computing instance creation method and device

    CN115361389A

  • Secure multi-tenant cloud subscription sharing

    US20220385629A1