D6.1.3: Final Operations Report including XSEDE Collaborative Services
Jules Wolfrat, Michael Rambadt · Zenodo (CERN European Organization for Nuclear Research) · 2015
This report presents the work performed by work package 6 of the PRACE-3IP project on the operation of the PRACE distributed infrastructure for Tier-0 and Tier-1 systems for the work extension in period July 2014 to end of January 2015. The main activity for this period has been the operation and coordination of the PRACE common services, which enable the interoperation among different PRACE HPC systems and which present a common interface to the PRACE Tier-0 and PRACE Tier-1 systems. Because of the limited available effort in this period for operational activities the number of new proposed changes has been limited to the inevitable ones, such as the removal of a system or a configuration change at a site. On operational security of the infrastructure, information on major vulnerabilities in basic system software has been exchanged and one incident at a partner site has been reported. Six Tier-0 systems have been operational for the project period under review: JUQUEEN at GCS@FZJ; CURIE at GENCI@CEA; SuperMUC at GCS@LRZ; Hermit at GCS@HLRS, upgraded to Hornet FERMI at CINECA; MareNostrum at BSC Hermit at GCS@HLRS, has been upgraded to a new system, Hornet, also at GCS@HLRS. With a peak performance of 3.8 PFlop/s for the new system this brings the total Tier-0 peak performance above 18 PFlop/s. Three Tier-1 systems have been upgraded in capacity, one has been replaced and one site has disappeared from the list. 23 Tier-1 sites are operational at the end of January 2015. Support for the enhancement of interoperability between PRACE and XSEDE has been given to three collaborating teams with members from both Europe and the US.