Virtual computing power pre-allocation method and device for intelligent computing center cloud platform

Abstract

The present invention provides a method and device for pre-allocating virtual computing power for an intelligent computing center cloud platform, and relates to the fields of intelligent computing centers, intelligent computing centers, computing power infrastructure, and intelligent computing cloud technology. The method comprises: step S1, receiving computing power resource application information, the computing power resource application information including type information of a computing power operation task and computing power resource requirement information of the computing power operation task; step S2, when the type information meets preset conditions, creating an operating environment for a container based on a preset amount of computing power resources; step S3, when a computing power operation task is received, allocating a first number of virtual accelerator cards to the container based on the type information and the computing power resource requirement information, the virtual accelerator cards being virtual accelerator cards of the intelligent computing center cloud platform, and being used to provide computing power resources; and step S4, executing the computing power operation task by the container based on the operating environment and the first number of virtual accelerator cards. The present invention greatly improves computing power resource utilization.

Description

Technical Field

[0001] The present invention relates to the fields of intelligent computing centers, smart computing centers, computing power infrastructure and intelligent computing cloud technology, and specifically to a method and device for pre-allocating virtual computing power for an intelligent computing center cloud platform. Background Art

[0002] With the rapid development of artificial intelligence technology, "intelligent computing centers" and "intelligent computing centers" have emerged.

[0003] An "Intelligent Computing Center" is a facility that utilizes large-scale heterogeneous computing resources, including general-purpose and intelligent computing power, to provide the computing power, data, and algorithms required for AI applications (such as AI deep learning model development, model training, and model inference). The Intelligent Computing Center encompasses facilities, hardware, and software, providing a full stack of capabilities, from bottom-level computing power to top-level application enablement.

[0004] “Intelligent Computing Center” includes but is not limited to “Smart Computing Center”.

[0005] "Intelligent Computing Center" refers to an artificial intelligence computing center. It is a type of computing power infrastructure that is based on artificial intelligence theory, adopts artificial intelligence computing architecture, and provides computing power services, data services, and algorithm services required for artificial intelligence applications.

[0006] "Computing power" is the core of "intelligent computing center" and "intelligent computing center". It is the ability of computer equipment or computing / data center to process information. It is the ability of computer hardware and software to work together to perform certain computing needs. It is the computing power to achieve target result output by processing information data. It is a new type of productivity that integrates information computing power, network carrying capacity, and data storage capacity. It mainly provides services to society through computing power infrastructure.

[0007] During model code development in an intelligent computing center, it's necessary to first create a container to run the model code and configure the runtime environment. The model code can then be imported into the container to facilitate user development. In existing technologies, even after the container is created and the runtime environment configured, there are cases where users develop the model code without actually using computing resources. In this case, the container occupies the computing resources of the intelligent computing center but does not actually perform any work, resulting in very low utilization of the computing resources of the intelligent computing center.

[0008] It can be seen that the existing technology has the problem of low utilization rate of computing resources in intelligent computing centers. Summary of the Invention

[0009] The present invention provides a method and device for pre-allocating virtual computing power of an intelligent computing center cloud platform, so as to solve the problem of low computing power resource utilization of the intelligent computing center in the prior art.

[0010] To solve the above problems, the present invention is achieved as follows:

[0011] In a first aspect, the present invention provides a method for pre-allocating virtual computing power of an intelligent computing center cloud platform, comprising:

[0012] Step S1: Receive computing power resource application information, where the computing power resource application information includes type information of the computing power operation task and computing power resource requirement information of the computing power operation task;

[0013] Step S2: If the type information meets the preset conditions, create a container operating environment based on the preset computing power resources;

[0014] Step S3: upon receiving the computing power operation task, allocating a first number of virtual accelerator cards to the container based on the type information and the computing power resource requirement information, where the virtual accelerator cards are virtual accelerator cards of the intelligent computing center cloud platform and are used to provide computing power resources;

[0015] Step S4: Execute the computing power operation task through the container based on the operating environment and the first number of virtual accelerator cards.

[0016] In one embodiment, the preset condition includes one of the following:

[0017] The type information indicates that the computing power running task is a long-term task;

[0018] The type information indicates that the computing power operation task is a step-growth type task;

[0019] The type information indicates that the computing power running task is an idle release type task.

[0020] In one embodiment, step S3 includes:

[0021] Step S31: Create a virtual mapping relationship table based on the computing power resource demand information. The virtual mapping relationship table is a relationship between the computing power running task and N virtual accelerator cards. The computing power resources provided by the N virtual accelerator cards match the computing power resource demand information, where N is a positive integer greater than or equal to 1.

[0022] Step S32: upon receiving the computing power operation task, determining the first number of virtual accelerator cards based on the type information and the computing power resource requirement information;

[0023] Step S33: Allocate the first number of virtual accelerator cards among the N virtual accelerator cards to the container based on the virtual mapping relationship table, where the first number of virtual accelerator cards are virtual accelerator cards that are in an idle state in the virtual mapping relationship table.

[0024] In one embodiment, step S32 includes:

[0025] Step S321: When the type information indicates that the computing power running task is a long-term task or an idle-release task, determine that the first number of virtual accelerator cards is the N virtual accelerator cards.

[0026] Step S322: When the type information indicates that the computing power running task is a step-growth type task, determine that the first number of virtual accelerator cards is M virtual accelerator cards, where M is a positive integer greater than or equal to 1, and M is less than or equal to N.

[0027] In one embodiment, the method further comprises:

[0028] Step S34: Set the first number of virtual accelerator cards in the virtual mapping relationship table to an occupied state.

[0029] In one embodiment, step S4 includes:

[0030] Step S41: When the type information indicates that the computing power operation task is a step-increase type task, obtain the load parameter of the container;

[0031] Step S42: When the load parameters of the container match the computing power resources provided by the first number of virtual accelerator cards, the computing power operation task is executed by the container based on the operating environment and the first number of virtual accelerator cards.

[0032] In one embodiment, the method further comprises:

[0033] Step S43: If the load parameters of the container do not match the computing power resources provided by the first number of virtual accelerator cards, allocate a second number of virtual accelerator cards to the container based on the computing power resource requirement information, where the second number is greater than the first number, and the computing power resources provided by the second number of virtual accelerator cards are less than or equal to the number of computing power resources required as indicated by the computing power resource requirement information.

[0034] Step S44: executing the computing power operation task through the container based on the operating environment and the second number of virtual accelerator cards;

[0035] Step S45: Set the second number of virtual accelerator cards in the virtual mapping relationship table to an occupied state.

[0036] In one embodiment, the method further comprises:

[0037] Step S46: Obtain usage parameters for providing the second number of virtual accelerator cards;

[0038] Step S47: If a usage parameter of a first virtual accelerator card among the second number of virtual accelerator cards indicates that the first virtual accelerator card has not been used within a first time period, release computing power resources provided by the first virtual accelerator card, and set the first virtual accelerator card in the virtual mapping relationship table to an idle state.

[0039] In one embodiment, step S3 includes:

[0040] Step S31′: monitoring the first number of virtual accelerator cards when the type information indicates that the computing power running task is an idle release type task;

[0041] Step S32′: when it is detected that a virtual accelerator card among the first number of accelerator cards is called, allocate the first number of virtual accelerator cards to the container.

[0042] In one embodiment, the method further comprises:

[0043] Step S5: When the type information indicates that the computing power running task is an idle release type task, monitoring the first number of virtual accelerator cards;

[0044] Step S6: When it is detected that the first number of virtual accelerator cards are not used within the second time period, the container is released, and the first number of virtual accelerator cards in the virtual mapping relationship table are set to an idle state.

[0045] In one embodiment, step S6 includes:

[0046] Step S61: When it is detected that the first number of virtual accelerator cards are not used within a second time period, data in the memories of the first number of virtual accelerator cards are saved to a storage space;

[0047] Step S62: Release the container, and set the first number of virtual accelerator cards in the virtual mapping relationship table to an idle state.

[0048] In a second aspect, the present invention further provides a virtual computing power pre-allocation device for an intelligent computing center cloud platform, comprising:

[0049] A receiving module is used to receive computing power resource application information, wherein the computing power resource application information includes type information of the computing power operation task and computing power resource requirement information of the computing power operation task;

[0050] A creation module, configured to create a container operating environment based on a preset amount of computing power resources when the type information meets preset conditions;

[0051] an allocation module, configured to allocate a first number of virtual accelerator cards to the container based on the type information and the computing power resource requirement information when receiving the computing power operation task, wherein the virtual accelerator cards are virtual accelerator cards of the intelligent computing center cloud platform and are used to provide computing power resources;

[0052] An execution module is used to execute the computing power operation task based on the operating environment and the first number of virtual accelerator cards through the container.

[0053] In a third aspect, the present invention also provides an electronic device comprising a processor, a memory, and a computer program stored in the memory and runnable on the processor. When the computer program is executed by the processor, the steps in the method for pre-allocating virtual computing power of the intelligent computing center cloud platform as described in the first aspect above are implemented.

[0054] In a fourth aspect, the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps in the virtual computing power pre-allocation method of the intelligent computing center cloud platform as described in the first aspect above are implemented.

[0055] In a fifth aspect, the present invention also provides a computer program product comprising computer instructions, which, when executed by a processor, implement the steps in the method for pre-allocating virtual computing power of the intelligent computing center cloud platform as described in the first aspect above.

[0056] In the present invention, step S1, receiving computing power resource application information, the computing power resource application information includes type information of computing power running tasks and computing power resource demand information of the computing power running tasks; step S2, when the type information meets the preset conditions, creating a container operating environment based on the preset computing power resource amount; step S3, when receiving the computing power running tasks, allocating a first number of virtual accelerator cards to the container based on the type information and the computing power resource demand information, the virtual accelerator card is a virtual accelerator card of the intelligent computing center cloud platform, and the virtual accelerator card is used to provide computing power resources; step S4, executing the computing power running tasks through the container based on the operating environment and the first number of virtual accelerator cards. In this way, creating the container operating environment through the preset computing power resource amount avoids the waste of computing power resources caused by occupying computing power resources when the container does not execute the computing power running tasks, thereby greatly improving the utilization rate of computing power resources; at the same time, executing the computing power running tasks through the container based on the operating environment and the first number of virtual accelerator cards avoids the waste of computing power resources caused by excessive allocation of virtual accelerator cards to the container, thereby greatly improving the utilization rate of computing power resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for the description of the present invention. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0058] Figure 1 This is a flow chart of a method for pre-allocating virtual computing power of an intelligent computing center cloud platform provided by the present invention;

[0059] Figure 2 This is a schematic diagram of pre-allocation of a virtual accelerator card provided by the present invention;

[0060] Figure 3 This is a structural diagram of a virtual computing power pre-allocation device for an intelligent computing center cloud platform provided by the present invention;

[0061] Figure 4 This is a structural diagram of an electronic device provided by the present invention. DETAILED DESCRIPTION

[0062] The following will clearly and completely describe the technical solutions of the present invention in conjunction with the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0063] The "computing power" mentioned in the present invention refers to: the ability of computer equipment or computing / data centers to process information, the ability of computer hardware and software to work together to execute certain computing requirements, and the computing power to achieve target result output by processing information data. It is a new type of productivity that integrates information computing power, network carrying capacity, and data storage capacity, and mainly provides services to society through computing power infrastructure.

[0064] The "computing power" (CP) mentioned in the present invention refers to: the ability of a data center server to process data and output results. It is a comprehensive indicator for measuring the computing power of a data center, including general computing power, supercomputing power and intelligent computing power. The commonly used unit of measurement is the number of floating-point operations performed per second (FLOPS, 1EFLOPS=10^18FLOPS). The larger the value, the stronger the comprehensive computing power. According to calculations, 1 EFLOPS is approximately the computing power output of 5 Tianhe-2A or 500,000 mainstream server CPUs or 2 million mainstream laptops. The calculation formula is: CP = CP 通用 +CP 智能 +CP 超级 .

[0065] The "Network Power" (NP) mentioned in this invention refers to: it is a manifestation of the data transmission capability of computing power facilities, including comprehensive capabilities such as network architecture, network bandwidth, transmission latency, intelligent management and scheduling, etc. It involves network transmission within and between data centers, and is a comprehensive indicator for measuring network transmission scheduling capabilities.

[0066] "Storage Power" (SP) as used in this document refers to the comprehensive capabilities of a data center in four areas: data storage capacity, performance, security and reliability, and environmental friendliness and low-carbon development. It serves as a comprehensive indicator of a data center's data storage capabilities, encompassing both external storage devices like storage arrays and internal server storage. Storage capacity is commonly measured in exabytes (EB, 1EB = 2^60 bytes), while performance is commonly measured in IOPS / TB (Input / Output Operations Per Second). Disaster recovery ratio is a key indicator of security and reliability.

[0067] The "computing power infrastructure" mentioned in the present invention refers to a new type of information infrastructure that integrates information computing power, network carrying capacity, and data storage capacity, and can realize the centralized calculation, storage, transmission and application of information.

[0068] The "new information infrastructure" mentioned in the present invention refers to: mainly including network infrastructure such as 5G networks, fiber-optic broadband networks, backbone networks, international communication networks, satellite Internet, computing power infrastructure such as data centers, general computing power centers, intelligent computing centers, supercomputing centers, and new technology facilities such as artificial intelligence, blockchain, and quantum computing.

[0069] The "computing power" mentioned in the present invention includes: general computing power, intelligent computing power and super computing power.

[0070] The "general computing power" mentioned in the present invention refers to the computing power provided by servers based on CPU (Central Processing Unit) chips, which is used to support basic general computing such as cloud computing and edge computing.

[0071] The "intelligent computing power" mentioned in this invention refers to: a computing platform based on specialized chips such as GPU (Graphics Processing Unit), FPGA (Field Programmable Gate Array), and ASIC (Application Specific Integrated Circuit) for various innovative artificial intelligence applications, such as natural language processing and machine vision.

[0072] The "supercomputing power" mentioned in the present invention refers to the computing power provided by high-performance computing clusters such as supercomputers. It utilizes the centralized computing resources of multiple computer systems working in parallel and uses a dedicated operating system to handle extremely complex or data-intensive problems. It is mainly used for calculations in cutting-edge scientific fields, such as planetary simulation, drug molecule design, genetic analysis, etc.

[0073] The term "intelligent computing center" as used in this document refers to a facility that utilizes large-scale heterogeneous computing resources, including general-purpose computing power (CPUs) and intelligent computing power (GPUs, FPGAs, ASICs, etc.), primarily to provide the computing power, data, and algorithms required for AI applications (such as AI deep learning model development, model training, and model inference). An intelligent computing center encompasses facilities, hardware, and software, providing a full stack of capabilities, from bottom-level computing power to top-level application enablement.

[0074] The "intelligent computing center cloud platform" mentioned in the present invention is referred to as "intelligent computing cloud", which refers to: a cloud computing platform that provides comprehensive services based on the hardware resources and software resources of the intelligent computing center.

[0075] The "intelligent computing center" mentioned in the present invention includes but is not limited to the "intelligent computing center".

[0076] The "intelligent computing center" mentioned in the present invention is an artificial intelligence computing center, which is a type of computing power infrastructure based on artificial intelligence theory, adopts artificial intelligence computing architecture, and provides computing power services, data services and algorithm services required for artificial intelligence applications.

[0077] The "computing power center" mentioned in the present invention refers to: a facility that is mainly composed of infrastructure such as wind, fire, water, electricity, and IT hardware and software equipment, and has computing power, transportation capacity, and storage capacity, including general data centers, intelligent computing centers, supercomputing centers, etc.

[0078] The "supercomputing center" mentioned in the present invention refers to: a supercomputing data center, which is a data center based on a supercomputer or a large-scale computing cluster, which can provide large-scale computing, storage and network services and other functions, and is widely used in application scenarios such as aerospace, national defense, oil exploration, climate modeling and genome sequencing.

[0079] The "computing resources" mentioned in the present invention refer to: technologies and facilities with information calculation, transmission, storage and application capabilities required for the development of a digital society, including but not limited to computing resources such as CPUs and GPUs, network resources such as switches and routers, storage resources such as storage arrays and distributed storage, security resources such as firewalls and intrusion detection systems, and supporting and guarantee resources such as wind, fire, water, and electricity.

[0080] The “model” mentioned in the present invention includes but is not limited to a “large language model” and a “multimodal large model”.

[0081] The "large language model" mentioned in this invention refers to a large language model (LLM), which is a language model with a large parameter scale. It is designed to understand and generate human language. It is trained with a large amount of text data and can perform a wide range of tasks including text summarization, translation, sentiment analysis, etc.

[0082] The "Multimodal Large Models" mentioned in this invention refer to models that combine multimodal information such as text, images, video, and audio for training, including but not limited to multimodal large language models.

[0083] The "computing power operation task" mentioned in the present invention refers to: a specific workload or job executed on computing power resources that requires a certain amount of computing power support, usually involving complex data processing, numerical calculations, model training or simulation scenarios.

[0084] The "virtual computing power" described in the present invention refers to the computing power resources provided by the virtual acceleration card of the intelligent computing center cloud platform. The virtual acceleration card can be obtained by dividing the physical acceleration card, and each virtual acceleration card provides part of the computing power resources of the physical acceleration card. For example, an 80G physical GPU acceleration card is divided into 4 virtual acceleration cards, and each virtual accelerator card can provide 20G computing power resources. The physical acceleration card is an acceleration card deployed by the intelligent computing center cloud platform, which can be a GPU accelerator card or a CPU accelerator card. The existing containerization technology and virtualization technology can be used to containerize and virtualize the physical acceleration card to obtain a virtual acceleration card of a certain specification.

[0085] "Pre-allocation" in this invention refers to pre-establishing a virtual mapping table between computing tasks and virtual accelerator cards. When a container needs to actually execute a computing task, the virtual accelerator card is actually assigned to the container based on the virtual mapping table. This pre-allocation prevents containers from occupying computing resources while waiting for computing tasks after creation, significantly improving computing resource utilization.

[0086] See Figure 1 , Figure 1 This is a flowchart of a method for pre-allocating virtual computing power of an intelligent computing center cloud platform provided by the present invention. Figure 1 As shown, the following steps are included:

[0087] Step S1: Receive computing power resource application information, where the computing power resource application information includes type information of the computing power running task and computing power resource requirement information of the computing power running task.

[0088] The above-mentioned computing power resource application information is the information sent by the user to the intelligent computing center cloud platform before calling the computing power resources of the intelligent computing center cloud platform to perform computing power operation tasks. Through the computing power resource application information, the intelligent computing center cloud platform can determine the maximum number of virtual acceleration cards required to provide computing power resources in the process of executing computing power operation tasks.

[0089] Among them, the computing power resource application information is the information corresponding to the computing power operation task, including the type information of the computing power operation task and the computing power resource demand information. The intelligent computing center cloud platform can determine the number of virtual acceleration cards allocated to the container based on the type information and computing power resource demand information.

[0090] The above type information represents the type of computing task. It should be noted that different types of computing tasks have different requirements for computing resources during execution. It is necessary to dynamically allocate computing resources for different types of computing tasks to reduce idle computing resources and improve computing resource utilization.

[0091] The computing power resource demand information represents the amount of computing power resources required to execute the computing power operation task. In some embodiments, the computing power resource demand information may include the number of virtual accelerator cards. In other embodiments, the computing power demand information may include the size of the required computing power resources. The intelligent computing center cloud platform may determine the number of virtual accelerator cards that are most required to execute the computing power operation task based on the size of the computing power resources.

[0092] Step S2: When the type information meets the preset conditions, create a container operating environment based on the preset computing power resources.

[0093] The above-mentioned preset conditions are used to determine whether the computing task requires pre-allocation, pre-configuring the operating environment with less computing resources. This operating environment is required for the container to execute the computing task. If the type information does not meet the preset conditions, pre-allocation is not performed. In this case, the container is directly created based on the computing resources corresponding to the computing resource requirement information and released after the computing task is completed. If the type information meets the preset conditions, the container's operating environment is created based on the preset computing resources.

[0094] It should be noted that the preset amount of computing power resources is a smaller amount of computing power resources that are pre-set. When creating the container's operating environment through the preset amount of computing power resources, there is no need to immediately allocate all the computing power resources corresponding to the computing power resource demand information. Instead, computing power resources are allocated when the container needs to execute computing power running tasks, thereby avoiding the waste of computing power resources caused by occupying computing power resources when the container does not execute computing power running tasks, and improving the utilization rate of computing power resources.

[0095] In some embodiments, the preset amount of computing power resources may be computing power resources provided by a CPU or a GPU. For example, the preset amount of computing power resources may be computing power resources provided by a 1GB CPU and 1GB memory. The image file is obtained from the image repository using the computing power resources provided by the 1GB CPU and loaded into the 1GB memory to create the container's operating environment.

[0096] In addition, in one embodiment, the preset condition includes one of the following:

[0097] The type information indicates that the computing power running task is a long-term task;

[0098] The type information indicates that the computing power operation task is a step-growth type task;

[0099] The type information indicates that the computing power running task is an idle release type task.

[0100] For example, the type of computing power running task can be a long-term type (that is, the computing power running task is a long-term type task). This type of computing power running task requires occupying computing power resources for a long time. Therefore, when allocating computing power resources to the container that executes this type of computing power running task, sufficient virtual accelerator cards can be directly allocated to the container based on the computing power resource demand information.

[0101] For example, the type of computing power running task can be a step-growth type (that is, the computing power running task is a step-growth type task). During the execution process, this type of computing power running task requires more computing power resources at some time and less computing power resources at other times. At this time, the container can be dynamically expanded or reduced according to the load of the container to avoid the container occupying too many computing power resources, thereby improving the utilization rate of computing power resources.

[0102] For example, the type of computing power running task can be an idle release type (the computing power running task is an idle release type task). The execution process of this type of computing power running task includes an execution stage and an idle stage. During the execution stage, the container executes the computing power running task based on the computing power resources corresponding to the computing power resource demand information. During the idle stage, the container does not use any computing power resources. At this time, the computing power resources occupied by the container can be released, and this part of the computing power resources can be allocated to other containers, thereby improving the utilization rate of the computing power resources.

[0103] Step S3: When receiving the computing power running task, a first number of virtual acceleration cards are allocated to the container based on the type information and the computing power resource demand information. The virtual acceleration cards are virtual acceleration cards of the intelligent computing center cloud platform, and the virtual acceleration cards are used to provide computing power resources.

[0104] Step S4: Execute the computing power operation task through the container based on the operating environment and the first number of virtual accelerator cards.

[0105] It should be noted that different types of computing power running tasks require different computing power resources at different times during the execution process. For different types of computing power running tasks, it is necessary to first determine the first number of virtual acceleration cards allocated to the container, and then use the container to execute the computing power running task based on the operating environment and the first number of virtual acceleration cards, so as to avoid the waste of computing power resources caused by too many virtual acceleration cards being allocated to the container, thereby greatly improving the utilization rate of computing power resources.

[0106] In the present invention, step S1, receiving computing power resource application information, the computing power resource application information includes type information of computing power running tasks and computing power resource demand information of the computing power running tasks; step S2, when the type information meets the preset conditions, creating a container operating environment based on the preset computing power resource amount; step S3, when receiving the computing power running tasks, allocating a first number of virtual accelerator cards to the container based on the type information and the computing power resource demand information, the virtual accelerator card is a virtual accelerator card of the intelligent computing center cloud platform, and the virtual accelerator card is used to provide computing power resources; step S4, executing the computing power running tasks through the container based on the operating environment and the first number of virtual accelerator cards. In this way, creating the container operating environment through the preset computing power resource amount avoids the waste of computing power resources caused by occupying computing power resources when the container does not execute the computing power running tasks, thereby greatly improving the utilization rate of computing power resources; at the same time, executing the computing power running tasks through the container based on the operating environment and the first number of virtual accelerator cards avoids the waste of computing power resources caused by excessive allocation of virtual accelerator cards to the container, thereby greatly improving the utilization rate of computing power resources.

[0107] In one embodiment, step S3 includes:

[0108] Step S31: Create a virtual mapping relationship table based on the computing power resource demand information. The virtual mapping relationship table is a relationship between the computing power running task and N virtual accelerator cards. The computing power resources provided by the N virtual accelerator cards match the computing power resource demand information, where N is a positive integer greater than or equal to 1.

[0109] Step S32: upon receiving the computing power operation task, determining the first number of virtual accelerator cards based on the type information and the computing power resource requirement information;

[0110] Step S33: Allocate the first number of virtual accelerator cards among the N virtual accelerator cards to the container based on the virtual mapping relationship table, where the first number of virtual accelerator cards are virtual accelerator cards that are in an idle state in the virtual mapping relationship table.

[0111] In the present invention, step S31 creates a virtual mapping relationship table based on the computing power resource demand information, wherein the virtual mapping relationship table is a relationship table between the computing power operation task and N virtual accelerator cards, and the computing power resources provided by the N virtual accelerator cards match the computing power resource demand information, where N is a positive integer greater than or equal to 1. Step S32 determines the first number of virtual accelerator cards based on the type information and the computing power resource demand information upon receiving the computing power operation task; and allocates the first number of virtual accelerator cards among the N virtual accelerator cards to the container based on the virtual mapping relationship table, wherein the first number of virtual accelerator cards are the virtual accelerator cards in the idle state in the virtual mapping relationship table. In this way, the first number of virtual accelerator cards is obtained from the N virtual accelerator cards through the virtual mapping relationship table and allocated to the container, so that the container can execute the computing power operation task based on the allocated first number of virtual accelerator cards.

[0112] The N virtual accelerator cards in the above virtual mapping relationship table are the maximum number of virtual accelerator cards required in the process of executing computing power running tasks. By establishing a virtual mapping relationship table between computing power execution tasks and N virtual accelerator cards, some or all of the N virtual accelerator cards can be allocated to the container that executes the computing power running tasks.

[0113] It should be noted that the N virtual accelerator cards may correspond to one or more physical accelerator cards.

[0114] Further, if Figure 2 As shown in the figure, in the process of creating a virtual mapping relationship table for computing power running tasks and N virtual accelerator cards, since different computing power running tasks have idle periods during execution, and virtual accelerator cards are not actually occupied during the idle periods, one virtual accelerator card can be pre-allocated to different computing power running tasks at the same time, thereby providing computing power services to more users. Figure 2 As shown, for computing power running task 1, the pre-allocated virtual accelerator cards are virtual accelerator card 1, virtual accelerator card 2 and virtual accelerator card 3; for computing power running task 2, the pre-allocated virtual accelerator cards are virtual accelerator card 3, virtual accelerator card 4 and virtual accelerator card 5, among which virtual accelerator card 3 is pre-allocated to two computing power running tasks at the same time, and the containers corresponding to the two computing power running tasks can occupy virtual accelerator card 3 to execute the computing power running tasks.

[0115] For example, the intelligent computing center cloud platform is equipped with 100 virtual accelerator cards. By pre-allocating some virtual accelerator cards and creating a virtual mapping relationship table with different computing power running tasks, it can provide computing power resources for computing power running tasks that require a total of 130 virtual accelerator cards, greatly improving the utilization of computing power resources.

[0116] The computing resources provided by the N virtual accelerator cards match the computing resource demand information. Specifically, the computing resources provided by the N virtual accelerator cards are greater than the computing resources required by the computing resource demand information. For example, if the computing resource demand information indicates a demand for 50GB of computing resources, and each virtual accelerator card provides a maximum of 20GB of computing resources, then three virtual accelerator cards are determined to provide the computing resources. The maximum computing resources that the three virtual accelerator cards can provide is 60GB, which is greater than the required 50GB of computing resources.

[0117] In some implementations, a preset percentage of over-allocated computing resources can be added. For example, if the preset percentage is set to 20%, and the computing resource demand information indicates a requirement for 35GB of computing resources, and each virtual accelerator card provides a maximum of 20GB of computing resources, the required computing resources after adding the preset percentage of over-allocated computing resources will be 42GB. In this case, computing resources will be determined to be provided by three virtual accelerator cards, avoiding the situation where inaccurate computing resource demand information in the computing resource application information leads to insufficient container computing resources.

[0118] In the above process of allocating the first number of virtual accelerator cards among the N virtual accelerator cards to the container based on the virtual mapping relationship table, it is necessary to first determine whether the first number of virtual accelerator cards among the N virtual accelerator cards are in an idle state. When the first number of virtual accelerator cards are idle, the first number of virtual accelerator cards are allocated to the container; if the number of virtual accelerator cards in the N virtual accelerator cards in the idle state is less than the first number, the allocation can be waited until the status of other virtual accelerator cards changes to an idle state, or the idle virtual accelerator cards can be directly allocated to the container, and the other virtual accelerator cards can be monitored after allocation to facilitate timely capacity expansion.

[0119] In one embodiment, step S32 includes:

[0120] Step S321: When the type information indicates that the computing power running task is a long-term task or an idle-release task, determine that the first number of virtual accelerator cards is the N virtual accelerator cards.

[0121] Step S322: When the type information indicates that the computing power running task is a step-growth type task, determine that the first number of virtual accelerator cards is M virtual accelerator cards, where M is a positive integer greater than or equal to 1, and M is less than or equal to N.

[0122] In the present invention, when the type information indicates that the computing power execution task is a long-duration task or an idle-release task, the first number of virtual accelerator cards is determined to be the N virtual accelerator cards; when the type information indicates that the computing power execution task is a step-increase task, the first number of virtual accelerator cards is determined to be M virtual accelerator cards, where M is a positive integer greater than or equal to 1 and less than or equal to N. In this way, different numbers of virtual accelerator cards are determined for different types of computing power execution tasks, avoiding the waste of computing power resources caused by allocating too many virtual accelerator cards to containers.

[0123] Among them, for long-term tasks or idle-release tasks, more computing power resources are usually required in the process of executing this type of computing power running tasks. Therefore, when allocating virtual accelerator cards, all virtual accelerator cards are directly allocated to avoid the situation where insufficient computing power resources lead to abnormal execution of computing power running tasks.

[0124] For tasks of the step-growth type, the computing resources required by the container change during the execution of this type of computing power running task. In this case, some virtual accelerator cards (i.e., M virtual accelerator cards) are first allocated out of N virtual accelerator cards, and then the capacity is expanded or reduced based on the actual usage of the container to avoid waste of computing power resources and abnormal execution of computing power running tasks due to insufficient computing power resources.

[0125] In some implementations, M virtual accelerator cards can be 1 virtual accelerator card, that is, a minimum number of virtual accelerator cards are set to perform computing power operation tasks, and the container is expanded when more computing power resources are required in the future, thereby maximizing the utilization of computing power resources.

[0126] In one embodiment, the method further comprises:

[0127] Step S34: Set the first number of virtual accelerator cards in the virtual mapping relationship table to an occupied state.

[0128] In the present invention, the first number of virtual accelerator cards in the virtual mapping relationship table are set to an occupied state, so that these virtual accelerator cards will not be occupied by other containers and facilitate subsequent expansion and contraction of containers based on the virtual mapping relationship table.

[0129] In one embodiment, step S4 includes:

[0130] Step S41: When the type information indicates that the computing power operation task is a step-increase type task, obtain the load parameter of the container;

[0131] Step S42: When the load parameters of the container match the computing power resources provided by the first number of virtual accelerator cards, the computing power operation task is executed by the container based on the operating environment and the first number of virtual accelerator cards.

[0132] When the type information indicates that the computing power running task is a step-growth type task, the computing power resources required by the container will change during the execution of this type of computing power running task. It is necessary to determine whether to scale up or down the container based on the load of the container.

[0133] Specifically, in the present invention, in step S41, when the type information indicates that the computing power operation task is a step-growth type task, the load parameters of the container are obtained; in step S42, when the load parameters of the container match the computing power resources provided by the first number of virtual accelerator cards, the computing power operation task is executed by the container based on the operating environment and the first number of virtual accelerator cards. In this way, when the load parameters of the container match the computing power resources provided by the first number of virtual accelerator cards, there is no need to expand or shrink the capacity, and the first number of virtual accelerator cards are directly used to execute the computing power operation task, thereby avoiding waste of computing power resources and abnormal execution of the computing power operation task due to insufficient computing power resources.

[0134] In one embodiment, the method further comprises:

[0135] Step S43: If the load parameters of the container do not match the computing power resources provided by the first number of virtual accelerator cards, allocate a second number of virtual accelerator cards to the container based on the computing power resource requirement information, where the second number is greater than the first number, and the computing power resources provided by the second number of virtual accelerator cards are less than or equal to the number of computing power resources required as indicated by the computing power resource requirement information.

[0136] Step S44: executing the computing power operation task through the container based on the operating environment and the second number of virtual accelerator cards;

[0137] Step S45: Set the second number of virtual accelerator cards in the virtual mapping relationship table to an occupied state.

[0138] In the present invention, step S43, when the load parameters of the container do not match the computing power resources provided by the first number of virtual accelerator cards, a second number of virtual accelerator cards is allocated to the container based on the computing power resource demand information, the second number is greater than the first number, and the computing power resources provided by the second number of virtual accelerator cards are less than or equal to the number of computing power resources required as represented by the computing power resource demand information; step S44, the computing power operation task is executed by the container based on the operating environment and the second number of virtual accelerator cards; step S45, the second number of virtual accelerator cards in the virtual mapping relationship table is set to an occupied state. In this way, when the load parameters of the container do not match the computing power resources provided by the first number of virtual accelerator cards, the container is expanded by allocating the second number of virtual accelerator cards to the container, so that the container can use sufficient computing power resources to execute the computing power operation task.

[0139] It should be noted that by setting the second number of virtual accelerator cards in the virtual mapping relationship table to an occupied state, the virtual accelerator card states in the virtual mapping relationship table are updated, and subsequent container expansion and contraction based on the virtual mapping relationship table is facilitated.

[0140] In some implementations, the second number of virtual accelerator cards may be determined as follows:

[0141] The N virtual accelerator cards are monitored by the monitoring layer of the intelligent computing center cloud platform;

[0142] In a case where a second virtual accelerator card is called among the N virtual accelerator cards, the second number of virtual accelerator cards are allocated to the container, where the second number of virtual accelerator cards includes the second virtual accelerator card and the first number of virtual accelerator cards, and the second virtual accelerator card is a virtual accelerator card among the N virtual accelerator cards excluding the first number of virtual accelerator cards.

[0143] The monitoring layer of the intelligent computing center cloud platform is used to monitor each virtual accelerator card in the intelligent computing center cloud platform. Specifically, the monitoring layer for monitoring the usage of the virtual accelerator card can be encapsulated on the driver layer of the virtual accelerator card, so that the calling status of each virtual accelerator card can be determined through the monitoring layer.

[0144] In one embodiment, the method further comprises:

[0145] Step S46: Obtain usage parameters for providing the second number of virtual accelerator cards;

[0146] Step S47: If a usage parameter of a first virtual accelerator card among the second number of virtual accelerator cards indicates that the first virtual accelerator card has not been used within a first time period, release computing power resources provided by the first virtual accelerator card, and set the first virtual accelerator card in the virtual mapping relationship table to an idle state.

[0147] It should be noted that since the computing power resources required for step-growth tasks are different in different time periods, there may be increases or decreases. Therefore, it is necessary not only to expand the container capacity but also to shrink the container capacity to avoid the situation where the container occupies computing power resources without using them, and to improve the utilization rate of computing power resources.

[0148] In the present invention, step S46 obtains usage parameters for the second number of virtual accelerator cards; and step S47, if a first virtual accelerator card among the second number of virtual accelerator cards is not used within a first time period, the computing resources provided by the first virtual accelerator card are released, and the first virtual accelerator card in the virtual mapping relationship table is set to an idle state. Thus, by determining the usage of each virtual accelerator card, if a first virtual accelerator card is not used within the first time period, the computing resources provided by the first virtual accelerator card are released, and the first virtual accelerator card is set to an idle state. This allows the first virtual accelerator card to be allocated to other containers to perform other computing tasks, thereby significantly improving the utilization of computing resources.

[0149] In addition, by setting the first virtual accelerator card in the virtual mapping relationship table to an idle state, when capacity expansion or contraction is performed subsequently, it can be determined that the first virtual accelerator card is in an idle state, and capacity expansion can be performed through the first virtual accelerator card.

[0150] In one embodiment, step S3 includes:

[0151] Step S31′: monitoring the first number of virtual accelerator cards when the type information indicates that the computing power running task is an idle release type task;

[0152] Step S32′: when it is detected that a virtual accelerator card among the first number of accelerator cards is called, allocate the first number of virtual accelerator cards to the container.

[0153] It should be noted that idle-release type tasks require the computing power resources provided by N virtual accelerator cards during the execution phase of the computing power running task, but do not require computing power resources during the idle phase. Therefore, it is necessary to determine the usage of the N virtual accelerator cards and expand and release them in a timely manner to improve the utilization of computing power resources.

[0154] In the present invention, in step S31', if the type information indicates that the computing power operation task is an idle release type task, the first number of virtual accelerator cards are monitored; in step S32', if it is detected that a virtual accelerator card among the first number of accelerator cards is called, the first number of virtual accelerator cards are allocated to the container. In this way, by monitoring the virtual accelerator cards to determine whether the container requires computing power resources to execute the computing power operation task, if a virtual accelerator card is called, it is considered that the container requires computing power resources to execute the computing power operation task. At this time, all virtual accelerator cards (i.e., the first number of virtual accelerator cards) can be allocated to the container, so that the container can execute the computing power operation task based on the computing power resources provided by the first number of virtual accelerator cards.

[0155] In one embodiment, the method further comprises:

[0156] Step S5: When the type information indicates that the computing power running task is an idle release type task, monitoring the first number of virtual accelerator cards;

[0157] Step S6: When it is detected that the first number of virtual accelerator cards are not used within the second time period, the container is released, and the first number of virtual accelerator cards in the virtual mapping relationship table are set to an idle state.

[0158] In the present invention, step S5, when the type information indicates that the computing power operation task is an idle release type task, the first number of virtual accelerator cards are monitored; step S6, when it is monitored that the first number of virtual accelerator cards are not used within the second time period, the container is released, and the first number of virtual accelerator cards in the virtual mapping relationship table are set to an idle state. In this way, by monitoring the first number of virtual accelerator cards to determine whether the container needs computing power resources to execute the computing power operation task, if the first number of virtual accelerator cards are not used within the second time period, it is considered that the container is in an idle stage and does not need computing power resources. At this time, the container is released, and the first number of virtual accelerator cards in the virtual mapping relationship table are set to an idle state, thereby avoiding the container occupying computing power resources for a long time without using them, thereby greatly improving the utilization rate of computing power resources.

[0159] In one embodiment, step S6 includes:

[0160] Step S61: When it is detected that the first number of virtual accelerator cards are not used within a second time period, data in the memories of the first number of virtual accelerator cards are saved to a storage space;

[0161] Step S62: Release the container, and set the first number of virtual accelerator cards in the virtual mapping relationship table to an idle state.

[0162] In the present invention, step S61, when it is detected that the first number of virtual accelerator cards are not used within the second time period, save the data in the memory of the first number of virtual accelerator cards to the storage space; step S62, release the container, and set the first number of virtual accelerator cards in the virtual mapping relationship table to an idle state. In this way, the data in the memory of the first number of virtual accelerator cards is first saved to the storage space, and then the container is released, and the first number of virtual accelerator cards in the virtual mapping relationship table are set to an idle state, so that the relevant data for executing the computing power operation task is all maintained in the storage space. When the container needs to execute the computing power operation task again in the future, the relevant data can be quickly obtained directly from the storage space, achieving rapid loading and improving the efficiency of executing the computing power operation task.

[0163] See Figure 3 , Figure 3 This is a structural diagram of a virtual computing power pre-allocation device for an intelligent computing center cloud platform provided by the present invention, such as Figure 3 As shown, the virtual computing power pre-allocation device 300 of the intelligent computing center cloud platform includes:

[0164] The receiving module 301 is configured to receive computing resource application information, wherein the computing resource application information includes type information of a computing task and computing resource requirement information of the computing task;

[0165] A creation module 302 is configured to create a container operating environment based on a preset amount of computing power resources when the type information meets a preset condition;

[0166] an allocation module 303 configured to allocate a first number of virtual accelerator cards to the container based on the type information and the computing resource requirement information when receiving the computing power operation task, wherein the virtual accelerator cards are virtual accelerator cards of the intelligent computing center cloud platform and are used to provide computing power resources;

[0167] The execution module 304 is used to execute the computing power operation task based on the operating environment and the first number of virtual accelerator cards through the container.

[0168] In one embodiment, the preset condition includes one of the following:

[0169] The type information indicates that the computing power running task is a long-term task;

[0170] The type information indicates that the computing power operation task is a step-growth type task;

[0171] The type information indicates that the computing power running task is an idle release type task.

[0172] In one embodiment, the allocation module 303 includes:

[0173] A creation submodule is configured to create a virtual mapping relationship table based on the computing power resource demand information, wherein the virtual mapping relationship table is a relationship between the computing power running task and N virtual accelerator cards, and the computing power resources provided by the N virtual accelerator cards match the computing power resource demand information, where N is a positive integer greater than or equal to 1;

[0174] a determination submodule, configured to, upon receiving the computing power operation task, determine the first number of virtual accelerator cards based on the type information and the computing power resource requirement information;

[0175] A first allocation submodule is configured to allocate a first number of the N virtual accelerator cards to the container based on the virtual mapping relationship table, where the first number of virtual accelerator cards are idle virtual accelerator cards in the virtual mapping relationship table.

[0176] In one embodiment, the determining submodule includes:

[0177] a first determining unit, configured to determine that the first number of virtual accelerator cards is the N virtual accelerator cards when the type information indicates that the computing power running task is a long-duration task or an idle-release task;

[0178] The second determination unit is used to determine that the first number of virtual accelerator cards is M virtual accelerator cards when the type information indicates that the computing power running task is a step-growth type task, where M is a positive integer greater than or equal to 1 and M is less than or equal to N.

[0179] In one embodiment, the virtual computing power pre-allocation device 300 of the intelligent computing center cloud platform further includes:

[0180] The first setting submodule is configured to set the first number of virtual accelerator cards in the virtual mapping relationship table to an occupied state.

[0181] In one embodiment, the execution module 304 includes:

[0182] A first acquisition submodule is configured to acquire a load parameter of the container when the type information indicates that the computing power running task is a step-increase type task;

[0183] The first execution submodule is configured to execute the computing power operation task based on the operating environment and the first number of virtual accelerator cards through the container when the load parameters of the container match the computing power resources provided by the first number of virtual accelerator cards.

[0184] In one embodiment, the virtual computing power pre-allocation device 300 of the intelligent computing center cloud platform further includes:

[0185] a second allocation submodule, configured to allocate a second number of virtual accelerator cards to the container based on the computing resource requirement information if the load parameters of the container do not match the computing resources provided by the first number of virtual accelerator cards, wherein the second number is greater than the first number, and the computing resources provided by the second number of virtual accelerator cards are less than or equal to the number of computing resources required as indicated by the computing resource requirement information;

[0186] A second execution submodule is configured to execute the computing power operation task based on the operating environment and the second number of virtual accelerator cards through the container;

[0187] The second setting submodule is configured to set the second number of virtual accelerator cards in the virtual mapping relationship table to an occupied state.

[0188] In one embodiment, the virtual computing power pre-allocation device 300 of the intelligent computing center cloud platform further includes:

[0189] A second acquisition submodule is configured to acquire usage parameters of the second number of virtual accelerator cards;

[0190] The first release submodule is configured to release the computing power resource device provided by the first virtual accelerator card and set the first virtual accelerator card in the virtual mapping relationship table to an idle state when a usage parameter of the first virtual accelerator card among the second number of virtual accelerator cards indicates that the first virtual accelerator card has not been used within a first time period.

[0191] In one embodiment, the allocation module 303 includes:

[0192] A monitoring submodule, configured to monitor the first number of virtual accelerator cards when the type information indicates that the computing power running task is an idle release type task;

[0193] The third allocation submodule is configured to allocate the first number of virtual accelerator cards to the container when it is detected that a virtual accelerator card among the first number of accelerator cards is called.

[0194] In one embodiment, the virtual computing power pre-allocation device 300 of the intelligent computing center cloud platform further includes:

[0195] A monitoring module, configured to monitor the first number of virtual accelerator cards when the type information indicates that the computing power running task is an idle release type task;

[0196] The release module is configured to release the container and set the first number of virtual accelerator cards in the virtual mapping relationship table to an idle state when it is detected that the first number of virtual accelerator cards are not used within a second time period.

[0197] In one embodiment, the release module comprises:

[0198] a saving submodule, configured to save the data in the memories of the first number of virtual accelerator cards to a storage space when it is detected that the first number of virtual accelerator cards are not used within a second time period;

[0199] The second releasing submodule is configured to release the container and set the first number of virtual accelerator cards in the virtual mapping relationship table to an idle state.

[0200] The virtual computing power pre-allocation device for the intelligent computing center cloud platform provided in the embodiment of the present invention is capable of implementing the various processes of the various embodiments of the virtual computing power pre-allocation method for the above-mentioned intelligent computing center cloud platform. The technical features correspond one to one and can achieve the same technical effects. To avoid repetition, they will not be described here.

[0201] It should be noted that the virtual computing power pre-allocation device of the intelligent computing center cloud platform in the embodiment of the present invention can be a device, or a component, integrated circuit, or chip in an electronic device.

[0202] The present invention also provides an electronic device, see Figure 4 , Figure 4 The electronic device includes a memory 401, a processor 402, and a program or instruction stored in the memory 401 and executed by the processor 402. Figure 1 Any steps in the corresponding embodiment of the virtual computing power pre-allocation method of the intelligent computing center cloud platform and the same beneficial effects are achieved will not be repeated here.

[0203] The processor 402 may be a CPU, an ASIC, an FPGA, or a GPU.

[0204] Those skilled in the art will appreciate that all or part of the steps of the embodiment of the method for pre-allocating virtual computing power for the intelligent computing center cloud platform can be accomplished through hardware related to program instructions, and the program can be stored in a readable medium.

[0205] The present invention also provides a readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above Figure 1Any steps in the corresponding embodiments of the method for pre-allocating virtual computing power for an intelligent computing center cloud platform can achieve the same technical effect and are not described here in detail to avoid repetition. The storage medium is, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0206] The present invention also provides a computer program product, comprising computer instructions, which, when executed by a processor, implement the above Figure 1 The various processes of the embodiment of the virtual computing power pre-allocation method of the corresponding intelligent computing center cloud platform can achieve the same technical effect. To avoid repetition, they will not be repeated here.

[0207] The terms "first", "second" and the like in the present invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. In addition, the terms "comprise" and "have" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or that are inherent to these processes, methods, products or devices. In addition, "and / or" is used in this application to represent at least one of the connected objects, for example A and / or B and / or C, which means comprising seven situations including single A, single B, single C, and both A and B exist, both B and C exist, both A and C exist, and both A, B and C exist.

[0208] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or apparatus comprising the element.

[0209] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better embodiment. Based on this understanding, the technical solution of this application, or the part that contributes to the existing technology, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal (which can be a mobile phone, computer, server, air conditioner, or second terminal device, etc.) to execute the methods of each embodiment of this application.

[0210] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of this application, ordinary technicians in this field can also make many forms without departing from the purpose of this application and the scope of protection of the claims, all of which are within the protection of this application.

Claims

1. A method for pre-allocating virtual computing power of an intelligent computing center cloud platform, characterized in that: include: Step S1: Receive computing power resource application information, where the computing power resource application information includes type information of the computing power operation task and computing power resource requirement information of the computing power operation task; Step S2: If the type information meets the preset conditions, create a container operating environment based on the preset computing power resources; Step S3: upon receiving the computing power operation task, allocating a first number of virtual accelerator cards to the container based on the type information and the computing power resource requirement information, where the virtual accelerator cards are virtual accelerator cards of the intelligent computing center cloud platform and are used to provide computing power resources; Step S4: executing the computing power operation task through the container based on the operating environment and the first number of virtual accelerator cards; The preset condition includes one of the following: The type information indicates that the computing power running task is a long-term task; The type information indicates that the computing power operation task is a step-growth type task; The type information indicates that the computing power running task is an idle release type task; The step S3 comprises: Step S31: Create a virtual mapping relationship table based on the computing power resource demand information. The virtual mapping relationship table is a relationship table between the computing power running task and N virtual accelerator cards. The computing power resources provided by the N virtual accelerator cards match the computing power resource demand information, where N is a positive integer greater than or equal to 1. Step S32: upon receiving the computing power operation task, determining the first number of virtual accelerator cards based on the type information and the computing power resource requirement information; Step S33: Allocate the first number of virtual accelerator cards among the N virtual accelerator cards to the container based on the virtual mapping relationship table, where the first number of virtual accelerator cards are virtual accelerator cards that are in an idle state in the virtual mapping relationship table. The step S32 includes: Step S321: When the type information indicates that the computing power running task is a long-term task or an idle-release task, determine that the first number of virtual accelerator cards is the N virtual accelerator cards. Step S322: When the type information indicates that the computing power running task is a step-growth type task, determine that the first number of virtual accelerator cards is M virtual accelerator cards, where M is a positive integer greater than or equal to 1, and M is less than or equal to N.

2. The method according to claim 1, wherein The method further comprises: Step S34: Set the first number of virtual accelerator cards in the virtual mapping relationship table to an occupied state.

3. The method according to claim 1 or 2, wherein: The step S4 comprises: Step S41: When the type information indicates that the computing power operation task is a step-increase type task, obtain the load parameter of the container; Step S42: When the load parameters of the container match the computing power resources provided by the first number of virtual accelerator cards, the computing power operation task is executed by the container based on the operating environment and the first number of virtual accelerator cards.

4. The method according to claim 3, wherein The method further comprises: Step S43: If the load parameters of the container do not match the computing power resources provided by the first number of virtual accelerator cards, allocate a second number of virtual accelerator cards to the container based on the computing power resource requirement information, where the second number is greater than the first number, and the computing power resources provided by the second number of virtual accelerator cards are less than or equal to the number of computing power resources required as indicated by the computing power resource requirement information. Step S44: executing the computing power operation task through the container based on the operating environment and the second number of virtual accelerator cards; Step S45: Set the second number of virtual accelerator cards in the virtual mapping relationship table to an occupied state.

5. The method according to claim 4, wherein The method further comprises: Step S46: Obtain usage parameters for providing the second number of virtual accelerator cards; Step S47: If a usage parameter of a first virtual accelerator card among the second number of virtual accelerator cards indicates that the first virtual accelerator card has not been used within a first time period, release computing power resources provided by the first virtual accelerator card, and set the first virtual accelerator card in the virtual mapping relationship table to an idle state.

6. The method according to claim 1 or 2, wherein: The step S3 comprises: Step S31′: monitoring the first number of virtual accelerator cards when the type information indicates that the computing power running task is an idle release type task; Step S32′: when it is detected that a virtual accelerator card among the first number of accelerator cards is called, allocate the first number of virtual accelerator cards to the container.

7. The method according to claim 1 or 2, wherein: The method further comprises: Step S5: When the type information indicates that the computing power running task is an idle release type task, monitoring the first number of virtual accelerator cards; Step S6: When it is detected that the first number of virtual accelerator cards are not used within the second time period, the container is released, and the first number of virtual accelerator cards in the virtual mapping relationship table are set to an idle state.

8. The method according to claim 7, wherein The step S6 comprises: Step S61: When it is detected that the first number of virtual accelerator cards are not used within a second time period, data in the memories of the first number of virtual accelerator cards are saved to a storage space; Step S62: Release the container, and set the first number of virtual accelerator cards in the virtual mapping relationship table to an idle state.

9. A virtual computing power pre-allocation device for an intelligent computing center cloud platform, characterized in that: include: A receiving module is used to receive computing power resource application information, wherein the computing power resource application information includes type information of the computing power operation task and computing power resource requirement information of the computing power operation task; A creation module, configured to create a container operating environment based on a preset amount of computing power resources when the type information meets preset conditions; an allocation module, configured to allocate a first number of virtual accelerator cards to the container based on the type information and the computing power resource requirement information when receiving the computing power operation task, wherein the virtual accelerator cards are virtual accelerator cards of the intelligent computing center cloud platform and are used to provide computing power resources; An execution module, configured to execute the computing power operation task based on the operating environment and the first number of virtual accelerator cards through the container; The preset condition includes one of the following: The type information indicates that the computing power running task is a long-term task; The type information indicates that the computing power operation task is a step-growth type task; The type information indicates that the computing power running task is an idle release type task; The allocation module includes: A creation submodule is configured to create a virtual mapping relationship table based on the computing power resource demand information, wherein the virtual mapping relationship table is a relationship between the computing power running task and N virtual accelerator cards, and the computing power resources provided by the N virtual accelerator cards match the computing power resource demand information, where N is a positive integer greater than or equal to 1; a determination submodule, configured to, upon receiving the computing power operation task, determine the first number of virtual accelerator cards based on the type information and the computing power resource requirement information; a first allocating submodule, configured to allocate a first number of virtual accelerator cards among the N virtual accelerator cards to the container based on the virtual mapping relationship table, where the first number of virtual accelerator cards are virtual accelerator cards that are in an idle state in the virtual mapping relationship table; The determination submodule includes: a first determining unit, configured to determine that the first number of virtual accelerator cards is the N virtual accelerator cards when the type information indicates that the computing power running task is a long-duration task or an idle-release task; The second determination unit is used to determine that the first number of virtual accelerator cards is M virtual accelerator cards when the type information indicates that the computing power running task is a step-growth type task, where M is a positive integer greater than or equal to 1 and M is less than or equal to N.

10. An electronic device, characterized in that: include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein when the program is executed by the processor, the steps of the method for pre-allocating virtual computing power of an intelligent computing center cloud platform as described in any one of claims 1 to 8 are implemented.

11. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the virtual computing power pre-allocation method for the intelligent computing center cloud platform according to any one of claims 1 to 8.

12. A computer program product, characterized in that It includes computer instructions, which, when executed by a processor, implement the steps of the virtual computing power pre-allocation method of the intelligent computing center cloud platform as described in any one of claims 1 to 8.

Images