20–23 Oct 2020
VBLHEP, JINR, Dubna, RUSSIA
Europe/Moscow timezone

Integration of distributed heterogeneous computing resources for the MPD experiment with DIRAC Interware

22 Oct 2020, 17:50
20m
https://meet3.interprocom.ru/RFBR_NICA

https://meet3.interprocom.ru/RFBR_NICA

Speaker

Igor Pelevanyuk (Joint Institute for Nuclear Research)

Description

RFBR grant: 18-02-40101
Computing and storage resources are essential for effective Monte-Carlo generation. In JINR there are several different resources. Tier1 and Tier2 clusters, Govorun supercomputer, JINR Cloud, and NICA cluster as computing resources. EOS disk storage and dCache tape storage. To use all these resources individually users have to be aware of many details and differences between resources and keep track of load on all of them. DIRAC Interware was adopted, configured, and expanded to fulfill requirements to allow massive simultaneous Monte-Carlo generation for MPD. For a year the joint infrastructure was used via DIRAC to run successfully 380000 jobs with an average duration of 5 hours each. The use of DIRAC allowed unified data access, performance estimation, and accounting across all resources.

Presentation materials