Resource allocation method based on task perception under heterogeneous memory architecture

Abstract

The invention discloses a resource allocation method based on task perception under heterogeneous memory architecture. The method is characterized by comprising the steps of process performance metadata recording, node task allocation recording, obtaining of a scheduling strategy of task characteristic perception, and obtaining of a migration strategy of page perception. Different tasks are distinguished, so that task allocation in all NUMA nodes is relatively uniform. Compared with a default task allocation strategy of a system, the cache contention of a system CPU and memory access competitions are relieved. Meanwhile, different read and write characteristic pages of the tasks are distinguished, and an adaptive placement strategy is adopted under the heterogeneous memory architecture, sothat the write operation frequency of an NVM is reduced, and the service life of the NVM is prolonged. By means of the method, most of write operations are generated in a DRAM so that the performanceloss can be reduced as much as possible.

Description

technical field [0001] The invention belongs to the technical field of computer memory management, and in particular relates to the application of novel non-volatile memory (NVM) and traditional dynamic random access memory (DRAM) to construct heterogeneous memory in widely used non-uniform memory access (NUMA) architecture servers, and On this basis, a method for efficient task resource allocation is realized through task characteristic perception. Background technique [0002] In September 1999, IBM integrated NUMA technology into IBM Unix. NUMA's breakthrough technology has completely broken away from the constraints of the traditional super-large bus on the multi-processing structure. It greatly enhances the number of processors, memory and I / O slots that a single operating system can manage. Faced with the current big data scenario, more and more applications are transformed from traditional computing-intensive applications to data-intensive applications. In order to ...

AI Technical Summary

Problems solved by technology

[0003] The purpose of the present invention is to propose a task-aware resource allocation method under the heterogeneous memory architecture. On the one hand, it adopts adaptive CPU and memory allocation for different types of applications; , using different page placement strategies to solve the defects of the existing NUMA management technology applied to heterogeneous memory, and to achieve efficient allocation of multi-tasks and efficient use of heterogeneous memory while ensuring low software overhead

Images