Method for automatically deploying operating environment along with operation for high-performance computing

Abstract

The invention discloses a method for automatically deploying an operating environment along with operation for high-performance computing. The method comprises the following steps: an isolated operating environment is preset at a high-performance computing system node; a user logs in an isolated operation environment of the login node; the user develops and debugs the operation and configures an operation running environment at the login node, and sends an operation running request to the operation management system; the job management system allocates computing nodes of high-performance computing system nodes for user jobs, and automatically deploys an upper-layer file system of a user to the allocated computing nodes synchronously along with the jobs while loading the jobs, and then corresponding computing nodes operates the jobs of the user; and after the job execution is finished, the computing node feeds back a result to the login node of the user, then the computing node removesthe own upper-layer file system, and the job process is finished. According to the invention, the job deployment efficiency of high-performance computing is improved, and the environmental configuration burden of a user is reduced while the burden of the user is reduced.

Description

technical field [0001] The invention relates to the field of computer system software, in particular to a high-performance computing-oriented method for automatically deploying an operating environment along with jobs. Background technique [0002] The current high-performance computer system divides physical nodes into login nodes and computing nodes. After the user sends a job request at the login node, the high-performance computing system allocates computing nodes through the job management system to respond to these requests and complete the corresponding tasks. task. In this process, currently there are mainly the following modes. [0003] 1. Traditional high-performance computing job execution methods [0004] In traditional high-performance computing systems, users share login nodes (such as figure 1 shown). After the user logs in to the login node, he develops and debugs his own application program in the environment of the login node. At this time, the user may...

AI Technical Summary

Problems solved by technology

A node configured by a user is likely to affect the use of other users, resulting in confusion in the configuration of the system environment, and such an environment cannot protect the privacy of users

[0008] 2. Job deployment methods that rely on containers

[0011] 1. Users use containers to encapsulate their own job running environment. The content of the container must contain the complete system environment required for the job to run, and these contents must be manually maintained by the user, which brings a greater burden to the user

[0012] 2. The user encapsulates the execution environment of the completed job and generates the container image. The user also needs to manually deploy the container image to the computing node or the global shared file system, and needs to specify the job management system to run the job when running the job through the job management system. Container image, these tasks still bring a large burden to users

[0013] 3. The isolation technology used by traditional containers brings a large performance overhead to the operation of the job

Images