Stable HPC Cluster Management Scheme through Performance Evaluation

HPC is representative tools for performing large-scale scientific calculation both academia and industry. Tachyon is a high-performance parallel computing system which constructed based on SUN Blade X6275. It composed of 3,200 computing nodes and infra-facilities. Also, this machine works with various software stacks such as file system, archive manager, compiler, debugger, parallel tools, etc. In this paper, we handle the requirements and requisites to build and manage an HPC cluster environment. In addition, we analyzed the history of batch job which include information performed by scheduler. By doing so, we are able to gauge the needs and performance of the next system to be introduced.