Unified managing and scheduling method of calculation resource in ARM architecture network cluster

Abstract

The present invention discloses a unified managing and scheduling method of a calculation resource in an ARM architecture network cluster, comprising the following steps: disposing a global container layer between a virtual layer and a physical layer, wherein the global container layer is used to intercept a system event between the virtual layer and the physical layer, and redirecting to a physical layer of a local or remote node; and adding a process corresponding to an application to an operating system kernel scheduling queue on a node in a cluster layer by using a configurable global scheduler, or deleting the process from the queue, wherein the configurable global scheduler does not change existing local resource scheduling in an operating system kernel. A single system mirror method is used, so that an existing application does not need to be modified, and calculation resources on multiple ARM servers may be fully used by using support of an existing Linux operating system for an SMP architecture, thereby improving operating performance, availability, and expandability of an application on an ARM server network cluster.

Description

technical field [0001] The invention belongs to a method for unified management and scheduling of computing resources, in particular to a method for unified management and dynamic scheduling of computing resources in a network cluster based on an ARM architecture. Background technique [0002] The industry has been looking for ARM-based system-level new products (SoC) as a substitute for Intel's x86-based chips in the data center. The British ARM company is only responsible for designing low-power chips, and then authorizes design manufacturers like Samsung and Qualcomm. Today, ARM chips have settled in the mobile device market, starting most of which are found in smartphones and tablets. Now, ARM is developing chips with 64-bit architecture and is looking for manufacturers to design this efficient SoC chip into data centers, which are more conducive to horizontal expansion, such as HPC, grid computing and data analysis. The server has attracted many chip manufacturers to ...

AI Technical Summary

Problems solved by technology

Existing single-node applications cannot directly take advantage of the convenience brought by these distributed computing frameworks, such as high availability (High Availability) and scalability (Scalability)

Images