Computer state updating method and device as well as distributed system

Abstract

The invention discloses a computer state updating method and device as well as a distributed system. The computer state updating method comprises the following steps of: receiving working state information reported by a proxy node of a computer; determining the working state expressed by the working state information as the current working state of the computer; comparing the current working state with the computer working state determined after receiving the last working state information; and when the comparison result meets the preset condition, generating the change information of the computer working state, and sending the change information to a master controller so as to update the computer working state recorded in the master controller. By applying the technical scheme, the existing between-group parallel and intra-group serial mode is changed into a between-group parallel and intra-group parallel mode, so that when the computer working state changes, the change information of the computer state is sent to the master controller in real time, and then the master controller updates the working state of the computer, thereby solving the problem that the working state of the computer can not be monitored and updated in real time in the existing technical scheme.

Description

technical field [0001] The present invention relates to the technical field of communication, and more specifically, to a computer state update method, device and distributed system. Background technique [0002] SLURM (Simple Linux Utility for Resource Management, simple linux resource management tool) system, as the most popular distributed system at present, is developed by the National Laboratory of LLNL (Lawrence Livermore National Laboratory) in the United States. The cluster system installed and deployed There are more than 2000. The SLURM system includes: a master controller and multiple computing node agents, each computer agent node is connected to a computer, and the master controller is responsible for monitoring all computer agent nodes in the system, thereby realizing the monitoring of all computers in the system. [0003] The real-time monitoring of the computer agent nodes by the general control adopts the method of serial within the group and parallel betwe...

AI Technical Summary

Problems solved by technology

However, when the working state of the computer in the SLURM system changes within 4 minutes, the master control cannot monitor and update the working state of the computer in time, for example: the working state of the first detected computer in the system changes from shutdown to startup, Since the general control adopts the method of serial monitoring within the group, only after the general control scans the working status of the computers in the same computer group, can it monitor the working status of the first detected computer again, and then when the computer is working When the state changes, the master control cannot monitor and update the working state of the computer in real time

Images