Resource scheduling method, device and system

Abstract

The invention provides a resource scheduling method. The method comprises the steps of obtaining a job program of a deep learning job, and converting the job program to obtain an intermediate representation of a calculation graph; segmenting the intermediate representation of the calculation graph to obtain a sub-graph set; packaging the sub-graph set to obtain workload mirror images correspondingto various accelerators; and determining a target accelerator from the accelerator cluster according to preset accelerator capability information, the service level condition submitted by the user and the information of the resource pool, and sending a corresponding workload mirror image to the target accelerator. According to the method, unified abstraction is carried out on operation programs of different frameworks by utilizing intermediate representation of the calculation graph; and based on the intermediate representation of the calculation graph, multiple workload mirror images, comprehensive accelerator capability information, service level conditions and information of a resource pool are obtained, a target accelerator is determined, corresponding workload mirror images are allocated to the target accelerator, accelerator resources are reasonably utilized, and the use efficiency is improved. The invention provides a resource scheduling device and system with the above beneficial effects.

Description

technical field [0001] The present application relates to the technical field of resource scheduling, and in particular to a resource scheduling method, a resource scheduling device, and a resource scheduling system. Background technique [0002] Computing-intensive workloads in deep learning place extremely high demands on the computing power of the computing system. In order to improve the computing performance of various workloads in deep learning, people usually use two methods to increase the computing power of the computing system. One is to increase the number of computer nodes participating in the calculation at the same time through distributed computing to obtain higher computing power, which is commonly referred to as horizontal expansion; the other is to increase the computing power of a single computer node, including the use of higher-performance central The processor (Center Processing Unit, CPU) and the use of dedicated computing accelerators, etc., are commo...

AI Technical Summary

Problems solved by technology

When using such a computing system, the resource usage efficiency of the resource scheduling method adopted by the current technology is poor, and heterogeneous accelerators cannot be scheduled, which has great limitations.

Images