A GPU capability scheduling management method and device

Through the CIM platform's cloud rendering streaming scheduling algorithm and monitoring and scheduling of server management module, dynamic allocation and load balancing of GPU resources are achieved, and the problem of low utilization of GPU resources in the existing technology is solved, and cloud rendering streaming service is improved efficiency and ability.

CN117519969BActive Publication Date: 2024-11-22浪潮智慧城市科技有限公司
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202311466141.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-11-07
Publication Date
2024-11-22
Estimated Expiration
2043-11-07

AI Technical Summary

Technical Problem

The existing GPU capability scheduling management methods rely on manual intervention, are inefficient and error-prone, and lack real-time monitoring and dynamic adjustment, resulting in low resource utilization and low task execution efficiency.

Method used

Through the CIM platform's cloud rendering and streaming scheduling algorithm, the server management module is used to monitor the usage status of GPU resources, and dynamic allocation and load balancing of resources are realized. The resource queue and resource comparison steps of the first-in-first-out principle are adopted to filter out idle resources and load them. Balance to ensure the rational use of GPU resources.

Benefits of technology

It improves the rational utilization and allocation of GPU resources, enhances the ability and number of cloud rendering and streaming services, solves the problem of underutilization of resources, and has advantages especially in multi-client access scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117519969B_ABST
    Figure CN117519969B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of GPU dynamic scheduling, and specifically provides a GPU capacity scheduling management method and device. With the help of the cloud rendering streaming scheduling algorithm of the CIM platform, the GPU resource usage status of each cloud rendering sub-node is monitored through the server management module of the platform. According to the scheduling algorithm provided by the platform, the resources are defined as two states, namely, idle and occupied, to achieve dynamic allocation and load balancing of cloud rendering streaming services. Compared with the prior art, the present invention can achieve reasonable utilization and allocation of GPU resource cluster rendering capabilities through this method, enhance the cloud rendering streaming service capabilities and quantity of the CIM platform, solve the problem that the cloud rendering streaming service allocation cannot fully utilize GPU resources, and has application advantages when multiple clients access the same resource scene.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of GPU dynamic scheduling, and specifically provides a GPU capability scheduling management method and device. Background Art

[0002] With the advancement of science and technology and the continuous development of industry technology, digital twins have ushered in new development opportunities. In order to reduce urban operating costs and improve efficiency, the city information model (CIM) has become a favorable tool for the development and intelligent management of each city. In order to better express the appearance of the city, cloud rendering technology has been developed, that is, high-voltage rendering is placed on the server side, allowing the client to load and run lightly, which occupies a certain proportion in platform construction.

[0003] Cloud rendering relies on the power of GPUs (graphics processing units). Due to the limited GPU resources, how to effectively manage and schedule GPU capabilities has become an important issue. Existing GPU capacity scheduling management methods mainly rely on manual intervention, which is inefficient and error-prone when dealing with complex GPU task allocation and scheduling problems. In addition, existing GPU capacity scheduling management methods usually lack real-time monitoring and dynamic adjustment of GPU device performance, resulting in low resource utilization and inefficient task execution. Summary of the invention

[0004] The present invention aims to solve the above-mentioned deficiencies in the prior art and provides a GPU capacity scheduling management method with strong practicability.

[0005] A further technical task of the present invention is to provide a GPU capacity scheduling management device that is rationally designed, safe and applicable.

[0006] The technical solution adopted by the present invention to solve its technical problem is:

[0007] A GPU capacity scheduling and management method uses the cloud rendering streaming scheduling algorithm of the CIM platform to monitor the GPU resource usage status of each cloud rendering sub-node through the server management module of the platform. According to the scheduling algorithm provided by the platform, the resources are defined as idle and occupied, so as to realize the dynamic allocation and load balancing of the cloud rendering streaming service.

[0008] Further, the following steps are included:

[0009] S1, client connection;

[0010] S2, resource comparison;

[0011] S3, scheduling processing;

[0012] S4, load balancing;

[0013] S5. Service disconnected.

[0014] Furthermore, in step S1, when the client connects to the cloud rendering streaming service, the CIM platform receives the connection instruction information and then stores the instruction in the resource queue. The resource queue follows the first-in-first-out principle and gives priority to the instruction that enters the queue first, waiting for the platform to perform resource comparison and resource scheduling.

[0015] Furthermore, in step S2, with the help of the capabilities of the CIM platform, the connection instruction is compared with resources in the resource pool, the flow path information of the streaming service of the resources in the resource pool is obtained, all resource flow paths are cached, and the resource queue is updated and marked as being processed.

[0016] Furthermore, in step S3, the usage status of each resource flow path is analyzed, the idle and available cloud rendering streaming services are screened out, the resource queue information is updated, and a logical judgment is made based on whether there are available cloud rendering services.

[0017] Furthermore, when there is an available cloud rendering flow path service, it is obtained and cached to provide a resource flow path list for the subsequent load balancing stage;

[0018] When there is no available cloud rendering pipeline service, the client is directly fed back with a prompt message indicating that there is no available service, and the queue information is updated.

[0019] Furthermore, in step S4, the CIM platform uses the capabilities of the server management module to achieve load balancing of cloud rendering streaming. According to the cloud rendering flow path cache list obtained through resource comparison, the cached cloud rendering services are traversed, and then based on the server node where each service is located, the GPU usage status corresponding to the service is obtained. After comparison and analysis, the cloud rendering flow path service with the least hardware resource pressure is returned to the client.

[0020] Furthermore, in step S4, if the client disconnects from the cloud rendering streaming service, the resource cache is modified, the resource status is modified, and the occupation status of the streaming path is released.

[0021] A GPU capability scheduling management device includes: at least one memory and at least one processor;

[0022] The at least one memory is used to store a machine-readable program;

[0023] The at least one processor is used to call the machine-readable program to execute a GPU capability scheduling management method.

[0024] Compared with the prior art, the GPU capacity scheduling management method and device of the present invention have the following outstanding beneficial effects:

[0025] The invention currently has more and more cloud rendering application scenarios, and the demand for GPU resources is greater. Through this method, the reasonable utilization and allocation of GPU resource cluster rendering capabilities can be achieved, and the cloud rendering streaming service capabilities and quantity of the CIM platform can be enhanced. The problem that GPU resources cannot be fully utilized in cloud rendering streaming service allocation is solved, and it has application advantages when multiple clients access the same resource scenario. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0027] Attached Figure 1 The present invention is a flowchart of a GPU capability scheduling management method. DETAILED DESCRIPTION

[0028] In order to enable those skilled in the art to better understand the solution of the present invention, the present invention is further described in detail below in conjunction with specific implementation methods. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0029] A best embodiment is given below:

[0030] like Figure 1 As shown, a GPU capacity scheduling management method in this embodiment uses the cloud rendering streaming scheduling algorithm of the CIM platform to monitor the GPU resource usage status of each cloud rendering sub-node through the server management module of the platform. According to the scheduling algorithm provided by the platform, the resources are defined as idle and occupied, so as to realize dynamic allocation and load balancing of cloud rendering streaming services.

[0031] The specific steps are as follows:

[0032] S1, client connection;

[0033] When the client connects to the cloud rendering streaming service, the CIM platform will receive the connection command information and then store the command in the resource queue. The resource queue follows the first-in-first-out principle and gives priority to the command that enters the queue first, waiting for the platform to compare and schedule resources.

[0034] S2, resource comparison;

[0035] With the help of the capabilities of the CIM platform, the connection instructions are compared with resources in the resource pool, the flow path information of the streaming service in the resource pool is obtained, all resource flow paths are cached, and the resource queue is updated and marked as being processed.

[0036] S3, scheduling processing;

[0037] Analyze the usage status of each resource flow path, filter out idle and available cloud rendering streaming services, update resource queue information, and make logical judgments based on whether there are available cloud rendering services;

[0038] There are cloud rendering pipeline services available;

[0039] If there is an available cloud rendering pipeline service, obtain and cache it to provide a resource pipeline list for the subsequent load balancing stage;

[0040] No cloud rendering pipeline service is available;

[0041] If there is no available cloud rendering pipeline service, the client will be directly fed back with a "no available service" prompt message and the queue information will be updated;

[0042] S4, load balancing;

[0043] With the help of the server management module on the CIM platform, load balancing of cloud rendering streaming is achieved. Based on the cloud rendering stream cache list obtained through resource comparison, the cached cloud rendering services are traversed, and then the GPU usage status of the service corresponding to each service is obtained based on the server node where the service is located. After comparison and analysis, the cloud rendering stream service with the least hardware resource pressure is returned to the client.

[0044] S5, service disconnection;

[0045] If the client disconnects from the cloud rendering streaming service, the resource cache and resource status will be modified to release the occupation status of the streaming path.

[0046] Based on the above method, a GPU capacity scheduling management device in this embodiment includes: at least one memory and at least one processor;

[0047] The at least one memory is used to store a machine-readable program;

[0048] The at least one processor is used to call the machine-readable program to execute a GPU capability scheduling management method.

[0049] The invention currently has more and more cloud rendering application scenarios, and the demand for GPU resources is greater. Through this method, the reasonable utilization and allocation of GPU resource cluster rendering capabilities can be achieved, and the cloud rendering streaming service capabilities and quantity of the CIM platform can be enhanced. The problem that GPU resources cannot be fully utilized in cloud rendering streaming service allocation is solved, and it has application advantages when multiple clients access the same resource scenario.

[0050] The above-mentioned specific implementations are only specific cases of the present invention. The patent protection scope of the present invention includes but is not limited to the above-mentioned specific implementations. Any technical solutions that comply with the claims of the present invention and any appropriate changes or substitutions made by ordinary technicians in the relevant technical field shall fall within the patent protection scope of the present invention.

[0051] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A GPU capacity scheduling management method, characterized in that: With the help of the cloud rendering streaming scheduling algorithm of the CIM platform, the server management module of the platform monitors the GPU resource usage status of each cloud rendering sub-node. According to the scheduling algorithm provided by the platform, the resources are defined as idle and occupied, realizing dynamic allocation and load balancing of cloud rendering streaming services. The steps are as follows: S1, client connection; When the client connects to the cloud rendering streaming service, the CIM platform receives the connection command information and then stores the command in the resource queue. The resource queue follows the first-in-first-out principle and prioritizes the command that enters the queue first, waiting for the platform to compare and schedule resources. S2, resource comparison; With the help of the CIM platform, the connection instruction is compared with resources in the resource pool, the flow path information of the streaming service in the resource pool is obtained, all resource flow paths are cached, and the resource queue is updated and marked as being processed; S3, scheduling processing; Analyze the usage status of each resource flow path, filter out idle and available cloud rendering streaming services, update resource queue information, and make logical judgments based on whether there are available cloud rendering services; When there is an available cloud rendering pipeline service, obtain and cache it to provide a resource pipeline list for the subsequent load balancing stage; When there is no available cloud rendering pipeline service, the client will be directly fed back with a prompt message indicating that there is no available service, and the queue information will be updated; S4, load balancing; S5. Service disconnected.

2. A GPU capacity scheduling management method according to claim 1, characterized in that: In step S4, the CIM platform uses the capabilities of the server management module to achieve load balancing of cloud rendering streaming. According to the cloud rendering flow path cache list obtained through resource comparison, the cached cloud rendering services are traversed, and then the GPU usage status corresponding to the service is obtained based on the server node where each service is located. After comparison and analysis, the cloud rendering flow path service with the least hardware resource pressure is returned to the client.

3. A GPU capacity scheduling management method according to claim 2, characterized in that: In step S4, if the client disconnects from the cloud rendering streaming service, the resource cache is modified, the resource status is modified, and the occupation status of the streaming path is released.

4. A GPU capacity scheduling management device, characterized in that: include: at least one memory and at least one processor; The at least one memory is used to store a machine-readable program; The at least one processor is configured to call the machine-readable program to execute the method according to any one of claims 1 to 3.