HPC resources integration at CMS - CERN Document Server

摘要

This document identifies minimal set of requirements on HPC based resources in order to run CMS workflows: it is meant to sketch the strategy for an effective exploitation of such machines. It describes possible ways to use HPC machines for any type of workflows, including data reconstruction and Monte Carlo digitisation and reconstruction, besides Monte Carlo generation and simulation, which would represent little value overall if considered alone. This needs the identification of the requirements in order to try to use HPC as any other traditional owned site, and thus dealing with large data input/output, stressing both storage and network.