Method for Aggregating Queue Information and Job Information in a Computing Environment

Abstract

A method for gathering queue information and job information in a computation environment is provided, wherein the computation environment has a configuration of three-layer multi-tree. This kind of configuration includes an main node as a root node, multiple sub-nodes as mid-layer nodes and multiple high performance computers as leaf-nodes, wherein the main node manages the entire computation environment including all sub-nodes and all high performance computers, and the sub-nodes manage the high performance computers in their precincts. Said method is characterized in that the method includes the following steps: a. the sub-nodes read the information of the high performance computers in their precincts; b. the sub-nodes periodically connect the high performance computers in their precincts according to the obtained information of the high performance computers, and use different commands to obtain the queue information and the job information based on the different types of the job management systems running in the high performance computers; c. the sub-nodes transform the obtained queue information and job information into specified format and reports them to the main node, so that the main node can obtain the current running state of the entire computation environment; d. the main node uses a dispatching strategy to auto-dispatch the jobs according to the current running state of the entire computation environment.

Description

technical field [0001] The present invention relates to the computing field, in particular to a method for aggregating queue information and job information in a computing environment, a system for running the method, and a computing environment using the queue information and job information aggregating method or system . Background technique [0002] At present, various high-performance computers are distributed geographically, and various high-performance computers use different operating systems, connection methods, job management systems, installed and optimized computing software, and their investment and maintenance costs High, only by fully improving the utilization efficiency of high-performance computers in the shared mode can the effectiveness of high-performance computer investment be fully utilized. Establishing a high-performance computing service environment (or computing environment) can achieve this goal by providing easy-to-use services to customers with c...

AI Technical Summary

Problems solved by technology

These grid software often have one or more disadvantages listed below: do not support one access device to connect multiple high-performance computers to the master node; do not support multiple access modes (such as SSH, RSH, etc.); no jobs It is inconvenient or unstable to obtain queue information and job status information; it does not support the acquisition, display and update of software information; it does not support automatic selection of job running queues; it does not support job migration and load balancing between high-performance computers, etc.

Images