Computing power application method of heterogeneous computing power chip, storage medium and related equipment

By collecting and integrating computing power parameter information of heterogeneous computing power chips, building a resource pool and matching business requests, the problem of low resource utilization of computing power chips is solved, and efficient resource reuse and waste reduction is achieved.

CN118964024APending Publication Date: 2024-11-15MIGU CO LTD +1
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202411085824.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-08
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

Computing power chips fail to make full use of all their computing power resources when processing services, resulting in a low resource utilization rate.

Method used

Collect computing power parameter information of multiple heterogeneous computing power chips, build computing power models and integrate them into computing power resource pools, match the target computing power model to process the request when receiving business requests, and improve resource utilization.

Benefits of technology

By reusing the computing power chip corresponding to the target computing power model, the utilization rate of computing power resources is improved and resource waste is reduced.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118964024A_ABST
    Figure CN118964024A_ABST
Patent Text Reader

Abstract

The invention provides a computing power application method of a heterogeneous computing power chip, a storage medium and related equipment, and relates to the technical field of computing power chips, the method comprises the following steps: collecting computing power parameter information of each heterogeneous computing power chip in a plurality of heterogeneous computing power chips, the structures of different heterogeneous computing power chips being different; constructing a computing power model of the corresponding heterogeneous computing power chip according to the computing power parameter information of each heterogeneous computing power chip so as to obtain a plurality of computing power models; integrating the plurality of computing power models to obtain a computing power resource pool; and when a service request is received, determining computing power demand information of the service request, matching a target computing power model from the computing power resource pool according to the computing power demand information, and processing the service request by using a computing power chip corresponding to the target computing power model. Therefore, the utilization rate of computing power resources is improved, and waste of the computing power resources is reduced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The embodiments of the present application relate to the technical field of computing power chips, and in particular to a computing power application method, storage medium, and related equipment of a heterogeneous computing power chip. Background Art

[0002] With the continuous development of computing chip technology, computing chips have played an increasingly important role in people's lives. When currently processing services, computing chips usually use computing resources exclusively for the services. However, due to the large variety of computing resources of computing chips, current computing chips do not use all types of computing resources of computing chips when processing services, resulting in low utilization of computing resources of computing chips. Summary of the invention

[0003] The embodiments of the present application provide a computing power application method, storage medium and related equipment of a heterogeneous computing power chip to solve the problem of low utilization of computing power resources of the computing power chip.

[0004] To solve the above problems, this application is implemented as follows:

[0005] In a first aspect, an embodiment of the present application provides a computing power application method of a heterogeneous computing power chip, including:

[0006] Collect the computing power parameter information of each heterogeneous computing power chip among multiple heterogeneous computing power chips. Different heterogeneous computing power chips have different structures.

[0007] Constructing a computing power model of a corresponding heterogeneous computing power chip according to the computing power parameter information of each heterogeneous computing power chip to obtain a plurality of computing power models, and integrating the plurality of computing power models to obtain a computing power resource pool;

[0008] When a service request is received, the computing power requirement information of the service request is determined, a target computing power model is matched from the computing power resource pool according to the computing power requirement information, and a computing power chip corresponding to the target computing power model is used to process the service request.

[0009] In a second aspect, an embodiment of the present application provides a computing power application device for a heterogeneous computing power chip, including:

[0010] A collection module is used to collect computing power parameter information of each heterogeneous computing power chip among multiple heterogeneous computing power chips. Different heterogeneous computing power chips have different structures.

[0011] A construction module is used to construct a computing power model of a corresponding heterogeneous computing power chip according to the computing power parameter information of each heterogeneous computing power chip to obtain a plurality of the computing power models; and integrate the plurality of the computing power models to obtain a computing power resource pool;

[0012] The processing module is used to determine the computing power requirement information of the business request when receiving the business request, match the target computing power model from the computing power resource pool according to the computing power requirement information, and use the computing power chip corresponding to the target computing power model to process the business request.

[0013] In a third aspect, an embodiment of the present application further provides an electronic device, comprising: a memory, a processor, and a program stored in the memory and executable on the processor; the processor is used to read the program in the memory to implement the steps in the method described in the first aspect above.

[0014] In a fourth aspect, an embodiment of the present application further provides a readable storage medium for storing a program, which, when executed by a processor, implements the steps in the method described in the first aspect above.

[0015] In a fifth aspect, an embodiment of the present application further provides a computer program product, comprising computer instructions, which, when executed by a processor, implement the steps in the method described in the first aspect above.

[0016] In an embodiment of the present application, computing power parameter information of each heterogeneous computing power chip in a plurality of heterogeneous computing power chips is collected, and the structures of different heterogeneous computing power chips are different; a computing power model of the corresponding heterogeneous computing power chip is constructed according to the computing power parameter information of each heterogeneous computing power chip to obtain a plurality of the computing power models; the plurality of computing power models are integrated to obtain a computing power resource pool; when a business request is received, the computing power requirement information of the business request is determined, and a target computing power model is matched from the computing power resource pool according to the computing power requirement information, and the computing power chip corresponding to the target computing power model is used to process the business request. In this way, when a business request is received, the computing power requirement information of the business request can be determined, and a target computing power model is matched from the computing power resource pool according to the computing power requirement information, and the computing power chip corresponding to the target computing power model is used to process the business request, that is, the computing power chip corresponding to the target computing power model can be reused to process the business request, thereby improving the utilization rate of the computing power resources of the computing power chip corresponding to the target computing power model and reducing the waste of computing power resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the description of the embodiments of the present application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0018] Figure 1 This is one of the flow charts of the computing power application method of the heterogeneous computing power chip provided in the embodiment of the present application;

[0019] Figure 2 This is the second flowchart of the computing power application method of the heterogeneous computing power chip provided in the embodiment of the present application;

[0020] Figure 3 It is a schematic diagram of an application scenario of a computing power application method of a heterogeneous computing power chip provided in an embodiment of the present application;

[0021] Figure 4 It is a structural schematic diagram of a computing power application device of a heterogeneous computing power chip provided in an embodiment of the present application;

[0022] Figure 5 It is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0023] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0024] The terms "first", "second" etc. in the embodiments of the present application are used to distinguish similar objects, and need not be used to describe a specific order or sequential order. In addition, the terms "include" and "have" and any variation thereof are intended to cover non-exclusive inclusions, for example, the process, method, system, product or equipment comprising a series of steps or units need not be limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or equipment. In addition, "and / or" is used in the present application to represent at least one of connected objects, such as A and / or B and / or C, representing the inclusion of separate A, separate B, separate C, and A and B all exist, B and C all exist, A and C all exist, and 7 situations that A, B and C all exist.

[0025] See also Figure 1 , Figure 1 1 is a flow chart of a method for applying computing power of a heterogeneous computing power chip provided in an embodiment of the present application. The embodiment of the present application can be applied to electronic devices. Figure 1 As shown, the computing power application method of the heterogeneous computing power chip may include the following steps:

[0026] Step 101: Collect computing power parameter information of each heterogeneous computing power chip among multiple heterogeneous computing power chips. Different heterogeneous computing power chips have different structures.

[0027] Among them, the specific type of the heterogeneous computing power chip is not limited here. Optionally, the heterogeneous computing power chip can be a computing power chip of X86 architecture or Advanced RISC Machine (ARM) architecture.

[0028] Among them, the computing power parameter information can be used to represent the computing power resources of each heterogeneous computing power chip. The computing power parameter information can include multi-dimensional parameter information, so that the computing power resources of the heterogeneous computing power chip can be described in multiple dimensions.

[0029] Step 102: construct a computing power model of the corresponding heterogeneous computing power chip according to the computing power parameter information of each heterogeneous computing power chip to obtain a plurality of the computing power models; integrate the plurality of the computing power models to obtain a computing power resource pool.

[0030] Among them, the computing power model can also be called a parameter model, and the above process of constructing the computing power model of the corresponding heterogeneous computing power chip according to the computing power parameter information of each heterogeneous computing power chip can also be called the computing power discovery process of the heterogeneous computing power chip.

[0031] Step 103: When a service request is received, determine the computing power requirement information of the service request, match the target computing power model from the computing power resource pool according to the computing power requirement information, and use the computing power chip corresponding to the target computing power model to process the service request.

[0032] Among them, the specific type of the service corresponding to the service request is not limited here. Optionally, the service can be a service for navigation route planning, optionally, the service can be a service for image editing, and optionally, the service can also be a service for video calls.

[0033] Among them, the target computing power model can be at least one of the multiple computing power models included in the computing power resource pool, and the specific number of the target computing power models is not limited here.

[0034] In an embodiment of the present application, through steps 101 to 103, computing power parameter information of each heterogeneous computing power chip in a plurality of heterogeneous computing power chips is collected, and the structures of different heterogeneous computing power chips are different; a computing power model of the corresponding heterogeneous computing power chip is constructed according to the computing power parameter information of each heterogeneous computing power chip to obtain a plurality of the computing power models; the plurality of computing power models are integrated to obtain a computing power resource pool; when a business request is received, the computing power requirement information of the business request is determined, and the target computing power model is matched from the computing power resource pool according to the computing power requirement information, and the computing power chip corresponding to the target computing power model is used to process the business request. In this way, when a business request is received, the computing power requirement information of the business request can be determined, and the target computing power model is matched from the computing power resource pool according to the computing power requirement information, and the computing power chip corresponding to the target computing power model is used to process the business request, that is, the computing power chip corresponding to the target computing power model can be reused to process the business request, thereby improving the utilization rate of the computing power resources of the computing power chip corresponding to the target computing power model and reducing the waste of computing power resources.

[0035] It should be noted that the computing resources corresponding to the multiple computing models included in the computing resource pool can be used to process services of different business requests, that is, the computing resources of the above-mentioned multiple computing models can be reused, thereby further improving the utilization rate of computing resources and further reducing the waste of computing resources.

[0036] As an optional implementation manner, the collecting computing power parameter information of each heterogeneous computing power chip in the plurality of heterogeneous computing power chips includes:

[0037] Collect computing power parameter information of each heterogeneous computing power chip among multiple heterogeneous computing power chips, and generate description information of computing power resources corresponding to the heterogeneous computing power chip according to the computing power parameter information of each heterogeneous computing power chip;

[0038] The step of integrating the plurality of computing power models to obtain a computing power resource pool includes:

[0039] Integrate the plurality of computing power models to obtain a computing power resource pool, and generate a computing power chip status list in the computing power resource pool;

[0040] Saving the description information to the computing power chip status list;

[0041] Among them, in the computing power chip status list, different heterogeneous computing power chips correspond to different identification information, and the description information includes the identification information.

[0042] Among them, the description information is used to describe the relevant information of the computing power resources corresponding to the heterogeneous computing power chip. For example, the description information may include at least one of the following information: identification information of the computing power resources corresponding to the heterogeneous computing power chip, utilization information, idle rate information, type information of the heterogeneous computing power chip, etc.

[0043] Among them, the implementation method of the present application can be called the computing power registration process of the computing power resources of the heterogeneous computing power chip, that is, the description information of the computing power resources of the heterogeneous computing power chip can be saved in the computing power chip status list, so that the statistical effect of the computing power resources of the heterogeneous computing power chip can be enhanced.

[0044] In the implementation manner of the present application, description information of the computing resources of each heterogeneous computing chip can be constructed based on the parameter information of each heterogeneous computing chip, and the description information can be saved in the computing chip status list. In this way, the management effect of multiple heterogeneous computing chips can be enhanced.

[0045] As an optional implementation, the computing power parameter information includes at least one of the following: the serial number (Serial Number, SN), Internet Protocol Address (Internet Protocol Address, IP) address, Media Access Control Address (Media Access Control Address, MAC) address, central processing unit (Central Processing Unit, CPU) information, memory information, disk information, graphics processing unit (Graphics Processing Unit, GPU) information and network processor (Neural Network Processing Unit, NPU) information of the heterogeneous computing power chip.

[0046] In the implementation manner of the present application, the diversity of types of computing power parameter information can be increased, and the more types of parameter information there are, the higher the accuracy of the description information of the computing power resources of the constructed heterogeneous computing power chip.

[0047] As an optional implementation, when receiving a service request, determining computing power requirement information of the service request, matching a target computing power model from the computing power resource pool according to the computing power requirement information, and processing the service request using a computing power chip corresponding to the target computing power model, including:

[0048] Monitor the computing power status of each computing power model in the computing power resource pool in real time, and obtain the computing power requirement information of the business request when receiving the business request;

[0049] Matching the target computing power model from the computing power resource pool according to the computing power demand information and the computing power status of each computing power model in the computing power resource pool;

[0050] The business request is processed using the computing power chip corresponding to the target computing power model.

[0051] Among them, the specific method of obtaining the computing power requirement information of the business request is not limited here. Optionally, the business request may carry computing power requirement information; optionally, the business request may carry storage address information corresponding to the computing power requirement information, and the above computing power requirement information can be obtained according to the storage address information, and the storage address indicated by the above storage address information can be located on the electronic device in the embodiment of the present application, and can also be located on other electronic devices, and the other electronic devices can be electrically connected to the electronic device in the embodiment of the present application.

[0052] Among them, the specific method of matching the target computing power model from the computing power resource pool according to the computing power demand information and the computing power status of each computing power model in the computing power resource pool is not limited here. Optionally, the computing power status of each computing power model in the computing power resource pool can be traversed to determine the target computing power model; optionally, the computing power status of the computing power chip corresponding to each computing power model in the computing power resource pool can be sorted according to the size of the computing power resources in the idle state. In this way, when the computing power demand information is determined, the target computing power model can be directly determined from multiple computing power models, thereby improving the efficiency of determining the target computing power model.

[0053] In the implementation manner of the present application, the computing power status of each computing power model in the computing power resource pool is monitored in real time, and when the business request is received, the computing power requirement information of the business request is obtained; the target computing power model is matched from the computing power resource pool according to the computing power requirement information and the computing power status of each computing power model in the computing power resource pool; and the business request is processed using the computing power chip corresponding to the target computing power model. In this way, the computing power of the computing power chip corresponding to the target computing power model allocated to the business request matches the computing power requirement corresponding to the computing power requirement information of the business request, so that the computing power of the computing power chip corresponding to the target computing power model is fully utilized, thereby improving the utilization rate of the computing power of the computing power chip corresponding to the target computing power model.

[0054] Among them, as an optional implementation method, using the computing power chip corresponding to the target computing power model to process business requests can be understood as calling the computing power resources of the computing power chip corresponding to the target computing power model to process business requests, that is, the computing power resources of the computing power chip corresponding to the target computing power model can be allocated to the business request to process the business request.

[0055] As another optional implementation, the processing of the service request by using a computing power chip corresponding to the target computing power model includes:

[0056] Creating a target container on a computing power chip corresponding to the target computing power model, wherein the target container is a container for processing the service request;

[0057] The service request is processed by the target container.

[0058] The specific type of the target container is not limited here. Optionally, the target container may be a container (docker), and the container (docker) may also be referred to as a docker container.

[0059] Among them, in the implementation manner of the present application, a target container is created on a computing power chip corresponding to a target computing power model, and a method of processing a business request through the target container can be referred to as computing power instantiation.

[0060] Optionally, when the target container is a docker container, that is, computing power instantiation can be understood as the process of creating a docker container on a computing power chip corresponding to the target computing power model with specified parameters, and the computing power demand corresponding to the computing power demand information can be mapped into specification parameters of the docker container, and the specification parameters of the docker container may include at least one of the following: the number of central processing unit (CPU) cores, the number of CPU threads, memory capacity, disk size, whether to use a graphics processing unit (GPU), whether to use a neural network processor (NPU), Internet Protocol Address (IP) (can be randomly assigned), and Media Access Control Address (MAC) (can be randomly assigned).

[0061] In the implementation manner of the present application, the service request is processed by the target container, so that the processing efficiency of the service request can be improved.

[0062] As another optional implementation, virtual machine technology can be used instead of docker containers. That is, in the process of scheduling computing power requirements, computing power instances (i.e., docker containers) can be created in advance. After obtaining computing power requirement information, computing power instances (i.e., docker containers) can be directly allocated according to the computing power requirement information, rather than creating docker containers in real time and then allocating them. In this way, the processing efficiency of docker containers for business requests can be improved.

[0063] It should be noted that, as an optional application scenario, the electronic device in the embodiment of the present application may include an edge management node and an agent component, and the agent component can be understood as a running program running on a computing chip. Specifically, the electronic device can send specification parameters to the agent component through the edge management node, and then the agent uses the docker container creation instruction to create a docker container. The electronic device can then store the created docker container in a database and allocate the docker container to a business request for use.

[0064] As an optional implementation manner, matching a target computing power model from the computing power resource pool according to the computing power demand information, and processing the service request using a computing power chip corresponding to the target computing power model, includes:

[0065] According to the computing power demand information and the computing power status of each computing power model in the computing power resource pool, a candidate computing power model is obtained by screening from the multiple computing power models included in the computing power resource pool, wherein the computing power chip corresponding to the candidate computing power model has a target computing power resource that meets the computing power demand information and has an idle computing power status;

[0066] If the target information of the candidate computing power model is verified, the candidate computing power model is determined as the target computing power model, and the computing power chip corresponding to the target computing power model is used to process the service request;

[0067] The target information includes at least one of the following:

[0068] The current computing power state of the computing power resources of the alternative computing power model is maintained in the idle state;

[0069] The computing power resources of the alternative computing power model include a computing power indicator that is less than a preset threshold.

[0070] Among them, the process of verifying the target information of the alternative computing power model can be referred to as the process of verifying whether the computing power state of the computing power chip corresponding to the alternative computing power model truly meets the computing power demand information, and the current computing power state of the computing power resources of the alternative computing power model maintained in an idle state can also be referred to as the online state of the computing power resources of the alternative computing power model in an idle state or the online state of the computing power resources of the alternative computing power model. The computing power indicators included in the computing power resources of the alternative computing power model are less than the preset threshold, which can be referred to as the utilization rate of the computing power indicators being lower than the preset threshold, and the above preset threshold is the system threshold.

[0071] It should be noted that the computing power resources of the alternative computing power model may include multiple computing power indicators, and the computing power indicators may be CPU utilization, GPU utilization, or NPU utilization, etc., which are not specifically limited here.

[0072] In the implementation manner of the present application, when the target information of the alternative computing power model is verified, the alternative computing power model is determined as the target computing power model, so that the accuracy of the target computing power model determination result can be further improved.

[0073] As an optional implementation manner, the processing the service request by using a computing power chip corresponding to the target computing power model includes:

[0074] Using the computing power chip corresponding to the target computing power model to process the service request, and switching the state information of the computing power chip corresponding to the target computing power model from an idle state to a used state;

[0075] When the computing power chip corresponding to the target computing power model completes processing of the service request, the state information of the computing power chip corresponding to the target computing power model is switched from the use state to the idle state.

[0076] It should be noted that when the computing power chip has description information, the description information may include status information, that is, the status information may be one of the description information. Alternatively, the status information may also be information different from the description information. No specific limitation is made here.

[0077] Among them, when the status information of the computing power chip corresponding to the target computing power model is an idle state, the first information can be used to represent the idle state. When the status information of the computing power chip corresponding to the target computing power model is a use state, the second information can be used to represent the use state. The first information and the second information are different. For example, the first information is 0 and the second information is 1. In this way, by identifying the first information and the second information, the status information of the computing power chip corresponding to the target computing power model can be accurately and efficiently determined.

[0078] In the implementation mode of the present application, when the computing power chip corresponding to the target computing power model completes the business request, the status information of the computing power chip corresponding to the target computing power model is switched from the use state to the idle state. In this way, the computing power chip corresponding to the target computing power model can be allocated to other businesses for use, thereby completing the timely recovery of the computing power resources of the computing power chip corresponding to the target computing power model and reducing the waste of computing power resources.

[0079] It should be noted that, as an optional application scenario, the electronic device in the embodiment of the present application may include a heterogeneous computing power management system, and the heterogeneous computing power management system may include a computing power resource management module, a computing power resource monitoring module and a computing power allocation scheduling module, and the collaboration steps of the above modules can be specifically referred to in Figure 2In this way, through the collaboration of the above modules, unified management of various computing powers of electronic devices and allocation and scheduling of computing power can be achieved, thereby improving the efficiency of computing resources.

[0080] For example, the computing power resource management module can abstract various heterogeneous (such as X86 architecture, ARM architecture) computing power chips, build computing power models, describe computing power resources from multiple dimensions, and thus build a unified computing power resource pool (i.e., a computing power resource pool obtained by integrating multiple computing power models). The agent program (agent) running on the computing power chip collects parameter information of the computing power chip (i.e., computing power parameter information) and builds a parameter model (i.e., computing power model) that describes the computing power chip resources. This process can be called computing power discovery, and the computing power parameter information can include at least one of the following: SN, IP address, MAC address, CPU information, memory, disk, GPU information, and NPU information. The parameter model after computing power discovery will be submitted by the agent to the edge management program (i.e., the program running on the edge management node). The edge management program will temporarily record the computing power chip information in the locally maintained computing power status table (i.e., the computing power chip status list), and then submit the parameter model to the heterogeneous management system. The system will verify the global uniqueness of the SN and MAC address in the parameters to ensure that each computing power chip has a unique identification (i.e., identification information), and then save the verified parameter model to the database. The above process can be called computing power registration.

[0081] It should be noted that the computing resource management module can perform benchmark tests on the computing resources of the NPU carried by each computing chip through the agent, specifically testing the number of tera operations per second (TOPS) that the NPU can perform, and associate the obtained NPU computing results with the computing chip. In addition, the computing resource management module can also provide detailed computing performance indicator descriptions for CPU, GPU, memory and disk on the parameter model.

[0082] Specifically, for CPU, the specific parameters include the number of CPU cores, number of threads, architecture platform and clock frequency (number of instructions executed per second), such as 8-core 16-thread ARM platform 2.5GHz; for GPU, the specific parameters include the number of GPU cores, core frequency and video memory capacity / type; for NPU, it is measured by the number of trillions of operations per second (TOPS), and this performance data can be obtained by running benchmark tests. The specific parameters for memory are capacity and frequency. The specific parameters for local disks are capacity and interface type. Through the computing resource management module, the underlying heterogeneous computing chips can be abstracted into a computing resource pool that can be allocated for use.

[0083] The computing resource monitoring module can monitor the computing resources of each computing chip in real time, that is, it can monitor the performance indicators and computing power usage status of each computing chip in real time. Specifically, the computing resource monitoring module can monitor the usage of the CPU, GPU, NPU memory, disk, network input / output (IO) of the computing chip, as well as the tasks being executed by the computing chip. Figure 3 The agent component can collect performance indicator data of computing power, including CPU usage, GPU usage, memory usage, NPU usage, network IO and other indicator data. The agent component reports the collected data to the edge management node through the intranet. The edge management program maintains the computing power status in the area, and computing power resource monitoring provides a basis for business computing power scheduling.

[0084] The computing power allocation and scheduling module can maintain the idle resources of the computing power chip. Maintaining the idle resources of the computing power chip is an indispensable step in resource allocation and scheduling. Only in this way can the electronic equipment clearly know which computing power chip resources can participate in the allocation and scheduling. For the schedulable computing power items, there are parameters to describe them, including: the computing power chip to which it belongs, the number of idle CPU cores, the number of idle CPU threads, the idle memory capacity, the idle disk capacity, the idle disk type, whether the NPU is idle, and whether the GPU is idle.

[0085] In addition, the computing power allocation and scheduling module can also perform computing power allocation and scheduling according to the computing power demand information of the business request, which can be specifically divided into two steps: computing power allocation and computing power recovery. The computing power allocation can be understood as the computing power allocation and scheduling module executing the steps of determining the computing power chip corresponding to the target computing power model and using the computing power chip corresponding to the target computing power model to process the business request in the embodiment of the present application. When executing the above steps, the steps of computing power resource search and computing power resource allocation can be executed in sequence. The computing power resource search is the process of screening the computing power chips (i.e., the computing power chips corresponding to the target computing power model) whose idle resources meet the computing power demand information in the resource pool according to the computing power demand information, and the computing power resource allocation is the process of allocating the computing power chips corresponding to the target computing power model to the business request for use. It should be noted that the computing power resources of the computing power chip corresponding to the target computing power model that has been used by the business request need to be removed from the idle resources (i.e., the status information of the computing power chip corresponding to the target computing power model can be switched to the use status). In addition, when the idle resources of a computing power chip are insufficient (that is, the computing power resources in the idle state are less than the preset value, and the preset value may be the computing power demand corresponding to the computing power demand information), it will no longer participate in the computing power allocation.

[0086] It should be noted that when determining the computing power chip corresponding to the target computing power model according to the computing power demand information, that is, when searching for computing power resources, the computing power chip corresponding to the target computing power model can be obtained through a variety of screening conditions, and optionally, at least one of the following can be included: the idle CPU core size range of the computing power chip, the idle CPU thread number size range, the idle memory size range, the idle disk size range, the idle disk type, whether the NPU is idle, and whether the GPU is idle. The above screening conditions can be used to screen out the computing power chip corresponding to the target computing power model that meets the computing power demand information.

[0087] Computing power recycling can be understood as: in the embodiment of the present application, when the computing power chip corresponding to the target computing power model completes the business request, the state of the computing power chip corresponding to the target computing power model is switched from the use state to the idle state, that is, computing power recycling is when the business request no longer needs computing power resources, calling the computing power recycling interface, performing the de-instantiation process, and returning the computing power resources used by the computing power instance to the idle resources of the computing power chip. Optionally, computing power de-instantiation is the process of actually deleting the docker container by calling the agent component through the edge management node.

[0088] See also Figure 4 , Figure 4 is a structural diagram of a computing power application device of a heterogeneous computing power chip provided in an embodiment of the present application, such as Figure 4 As shown, the computing power application device 400 of the heterogeneous computing power chip includes:

[0089] The collection module 401 is used to collect computing power parameter information of each heterogeneous computing power chip among multiple heterogeneous computing power chips, and the structures of different heterogeneous computing power chips are different;

[0090] A construction module 402 is used to construct a computing power model of a corresponding heterogeneous computing power chip according to the computing power parameter information of each heterogeneous computing power chip to obtain a plurality of computing power models; and integrate the plurality of computing power models to obtain a computing power resource pool;

[0091] Processing module 403 is used to determine the computing power requirement information of the business request when receiving the business request, match the target computing power model from the computing power resource pool according to the computing power requirement information, and use the computing power chip corresponding to the target computing power model to process the business request.

[0092] As an optional implementation, the collection module 401 is further used to collect computing power parameter information of each heterogeneous computing power chip in the multiple heterogeneous computing power chips, and generate description information of computing power resources corresponding to the heterogeneous computing power chip according to the computing power parameter information of each heterogeneous computing power chip;

[0093] Building block 402 includes:

[0094] An integration submodule, used to integrate the plurality of computing power models to obtain a computing power resource pool, and generate a computing power chip status list in the computing power resource pool;

[0095] A saving submodule, used to save the description information into the computing power chip status list;

[0096] Among them, in the computing power chip status list, different heterogeneous computing power chips correspond to different identification information, and the description information includes the identification information.

[0097] As an optional implementation, the computing power parameter information includes at least one of the following: the serial number, Internet Protocol address, media access control address, central processing unit information, memory information, disk information, graphics processor information and network processor information of the heterogeneous computing power chip.

[0098] As an optional implementation, the processing module 403 includes:

[0099] An acquisition submodule, used to monitor the computing power status of each computing power model in the computing power resource pool in real time, and to obtain the computing power requirement information of the business request when receiving the business request;

[0100] A matching submodule, configured to match the target computing power model from the computing power resource pool according to the computing power demand information and the computing power status of each computing power model in the computing power resource pool;

[0101] The processing submodule is used to process the business request using the computing power chip corresponding to the target computing power model.

[0102] As an optional implementation, the processing submodule includes:

[0103] A creating unit, configured to create a target container on a computing power chip corresponding to the target computing power model, wherein the target container is a container for processing the service request;

[0104] The first processing unit is configured to process the service request through the target container.

[0105] As an optional implementation, the matching submodule is further used to screen out an alternative computing power model from the multiple computing power models included in the computing power resource pool according to the computing power demand information and the computing power state of each computing power model in the computing power resource pool, and the computing power chip corresponding to the alternative computing power model has a target computing power resource that meets the computing power demand information and has an idle computing power state;

[0106] The processing submodule is further configured to determine the candidate computing power model as the target computing power model when the target information of the candidate computing power model is verified, and use the computing power chip corresponding to the target computing power model to process the service request;

[0107] The target information includes at least one of the following:

[0108] The current computing power state of the computing power resources of the alternative computing power model is maintained in the idle state;

[0109] The computing power resources of the alternative computing power model include a computing power indicator that is less than a preset threshold.

[0110] As an optional implementation, the processing submodule includes:

[0111] A processing unit, configured to process the service request using a computing power chip corresponding to the target computing power model, and switch the state information of the computing power chip corresponding to the target computing power model from an idle state to a used state;

[0112] The switching submodule is used to switch the state information of the computing power chip corresponding to the target computing power model from the use state to the idle state when the computing power chip corresponding to the target computing power model completes processing of the business request.

[0113] The computing power application device 400 of the heterogeneous computing power chip can realize the embodiment of the present application Figure 1 The various processes of the method embodiment and the same beneficial effects are achieved, and will not be described again here to avoid repetition.

[0114] The present application also provides an electronic device. Figure 5 , the electronic device may include a processor 501, a memory 502, and a program 5021 stored in the memory 502 and executable on the processor 501. When the program 5021 is executed by the processor 501, Figure 1 Any steps in the corresponding method embodiments and achieving the same beneficial effects will not be repeated here.

[0115] The present application also provides a computer program product, including computer instructions, which can achieve the above-mentioned Figure 1 Any steps in the corresponding method embodiments can achieve the same technical effect and will not be repeated here to avoid repetition.

[0116] A person skilled in the art can understand that all or part of the steps of implementing the above-mentioned embodiment method can be completed by hardware related to program instructions, and the program can be stored in a readable medium. The present application also provides a readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above-mentioned Figure 1 Any steps in the corresponding method embodiments can achieve the same technical effect and will not be repeated here to avoid repetition.

[0117] The storage medium is, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0118] The above is a preferred implementation of the embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles described in the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A computing power application method for heterogeneous computing power chips, It is characterized in that include: Collect the computing power parameter information of each heterogeneous computing power chip among multiple heterogeneous computing power chips. Different heterogeneous computing power chips have different structures. Constructing a computing power model of a corresponding heterogeneous computing power chip according to the computing power parameter information of each heterogeneous computing power chip to obtain a plurality of computing power models, and integrating the plurality of computing power models to obtain a computing power resource pool; When a service request is received, the computing power requirement information of the service request is determined, a target computing power model is matched from the computing power resource pool according to the computing power requirement information, and a computing power chip corresponding to the target computing power model is used to process the service request.

2. The method according to claim 1, It is characterized in that The collecting of computing power parameter information of each heterogeneous computing power chip in the plurality of heterogeneous computing power chips includes: Collect computing power parameter information of each heterogeneous computing power chip among multiple heterogeneous computing power chips, and generate description information of computing power resources corresponding to the heterogeneous computing power chip according to the computing power parameter information of each heterogeneous computing power chip; The step of integrating the plurality of computing power models to obtain a computing power resource pool includes: Integrate the plurality of computing power models to obtain a computing power resource pool, and generate a computing power chip status list in the computing power resource pool; Saving the description information to the computing power chip status list; Among them, in the computing power chip status list, different heterogeneous computing power chips correspond to different identification information, and the description information includes the identification information.

3. The method according to claim 2, It is characterized in that The computing power parameter information includes at least one of the following: the serial number, Internet Protocol address, media access control address, central processing unit information, memory information, disk information, graphics processor information and network processor information of the heterogeneous computing power chip.

4. The method according to any one of claims 1 to 3, It is characterized in that When receiving a service request, determining computing power requirement information of the service request, matching a target computing power model from the computing power resource pool according to the computing power requirement information, and processing the service request using a computing power chip corresponding to the target computing power model, including: Monitor the computing power status of each computing power model in the computing power resource pool in real time, and obtain the computing power requirement information of the business request when receiving the business request; Matching the target computing power model from the computing power resource pool according to the computing power demand information and the computing power status of each computing power model in the computing power resource pool; The business request is processed using the computing power chip corresponding to the target computing power model.

5. The method according to claim 4, It is characterized in that The processing of the service request by using the computing power chip corresponding to the target computing power model includes: Creating a target container on a computing power chip corresponding to the target computing power model, wherein the target container is a container for processing the service request; The service request is processed by the target container.

6. The method according to claim 4, It is characterized in that The matching the target computing power model from the computing power resource pool according to the computing power demand information and the computing power status of each computing power model in the computing power resource pool includes: According to the computing power demand information and the computing power status of each computing power model in the computing power resource pool, a candidate computing power model is obtained by screening from the multiple computing power models included in the computing power resource pool, wherein the computing power chip corresponding to the candidate computing power model has a target computing power resource that meets the computing power demand information and has an idle computing power status; The processing of the service request by using the computing power chip corresponding to the target computing power model includes: If the target information of the candidate computing power model is verified, the candidate computing power model is determined as the target computing power model, and the computing power chip corresponding to the target computing power model is used to process the service request; The target information includes at least one of the following: The current computing power state of the computing power resources of the alternative computing power model is maintained in the idle state; The computing power resources of the alternative computing power model include a computing power indicator that is less than a preset threshold.

7. The method according to claim 4, It is characterized in that The processing of the service request by using the computing power chip corresponding to the target computing power model includes: Using the computing power chip corresponding to the target computing power model to process the service request, and switching the state information of the computing power chip corresponding to the target computing power model from an idle state to a used state; When the computing power chip corresponding to the target computing power model completes processing of the service request, the state information of the computing power chip corresponding to the target computing power model is switched from the use state to the idle state.

8. An electronic device, include: A memory, a processor, and a program stored in the memory and executable on the processor; characterized in that the processor is used to read the program in the memory to implement the steps in the computing power application method of the heterogeneous computing power chip as described in any one of claims 1 to 7.

9. A readable storage medium for storing a program, It is characterized in that When the program is executed by the processor, the steps in the computing power application method of the heterogeneous computing power chip as described in any one of claims 1 to 7 are implemented.

10. A computer program product, It is characterized in that It includes computer instructions, which, when executed by a processor, implement the steps in the computing power application method of a heterogeneous computing power chip as described in any one of claims 1 to 7.

Citation Information

Patent Citations

  • Task processing method and device applied to heterogeneous computing power system

    CN114035943A

  • Computing power unloading method and device, electronic equipment and storage medium

    CN115509644A

  • Computing power network task scheduling method and device, computer equipment and storage medium

    CN116263701A

  • Heterogeneous computing power equipment selection method, device, equipment, medium, product and system

    CN118245238A