Resource scheduling method and device, equipment and storage medium
CN122924974APending Publication Date: 2026-10-16GUANGDONG OPPO MOBILE TELECOMMUNICATIONS CORP LTD
0 Cites 0 Cited by
Patent Information
- Application Number
- CN202510430207.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-07
- Publication Date
- 2026-10-16
AI Technical Summary
Technical Problem
[0003]由于不同推理服务对GPU的需求各不相同，且线上的推理工作负载随着时间和用户数等因素的变动对资源的需求也是动态变化的，如何高效的为各类AI推理服务分配适合的GPU资源，将影响AI数据中心GPU资源的利用率和能效开销
Benefits of technology
[0019]When a user needs to add a first inference service, the computer equipment can test the GPU resource requirements of the first inference service and schedule GPU resources for the first inference service based on the test results. This method can be applied to GPU resource scheduling in AI data center inference clusters. While ensuring the operational needs of inference services, it can more rationally schedule and allocate GPU resources in the inference cluster for each inference service, effectively improving the resource utilization of the inference service cluster and saving energy costs.
✦ Generated by Eureka AI based on patent content.
Abstract
The application discloses a resource scheduling method and device, equipment and a storage medium, and belongs to the field of artificial intelligence. The method is executed by a computer device, and the computer device comprises at least two graphic processing units (GPUs). The method comprises the following steps: receiving a deployment request of a first inference service, wherein the deployment request is used for requesting to add the first inference service; testing a GPU resource requirement condition of the first inference service to obtain a test result, wherein the test result is used for indicating GPU resources required for running the first inference service; and scheduling a first GPU from the at least two GPUs to deploy the first inference service according to the test result and usage conditions of the at least two GPUs, wherein the first GPU comprises at least one of the following: an independent GPU and a shared GPU obtained by cutting an independent GPU. The method can accurately and efficiently schedule GPU resources for inference services.
Need to check novelty before this filing date? Find Prior Art