Parallel computing resource allocation method and device

Abstract

The invention discloses a parallel computing resource allocation method and device. The method includes receiving a current resource allocation request, thereby obtaining the total number of resources of the current resource allocation request; obtaining and calculating resource-related information contained in each cluster node in a cluster; according to the resource-related information of each cluster node, calculating resource allocation price of each cluster node; and according to the resource allocation price of each cluster node and the total number of resources of the current resource allocation request, performing resource allocation on a request task corresponding to the current resource allocation request. Through the parallel computing resource allocation method, the utilization rate of the entire cluster resource can be effectively improved.

Description

technical field [0001] The invention relates to the field of cluster resource management and parallel computing, in particular to a parallel computing resource allocation method and device. Background technique [0002] In recent years, parallel computing clusters have been more and more used in artificial intelligence-related fields such as speech recognition, image recognition, and natural language understanding, which has significantly improved the task processing capabilities of these fields, especially the deep learning tasks in these fields. The processing capability has been greatly improved; however, as the amount of data processing increases, a single resource is often unable to carry complex computing tasks, which requires more and more resources in the parallel computing cluster to perform corresponding tasks to meet the actual application requirements, such as using multiple A GPU (Graphic Processing Unit, GPU) forms a GPU cluster to perform the computing require...

AI Technical Summary

Problems solved by technology

[0003] The existing parallel computing cluster resource allocation method generally simply allocates resources in the cluster as an auxiliary resource other than CPU, memory, and hard disk. When cluster nodes allocate resources, they generally use mutually exclusive and exclusive methods to allocate Resources are allocated, that is, when the idle resources in the cluster nodes meet the task requirements, they will be allocated, and the allocated resources will monopolize the resources until the task is completed. However, this resource allocation method will cause the task to wait for a long time , cluster parallel processing capability decline and low resource utilization, etc.

Images