D6.3: Second Annual Operations Report of the Tier-1 Service
Jules Wolfrat · Zenodo (CERN European Organization for Nuclear Research) · 2013
The primary objective of WP6 ‘European HPC Infrastructure Operation and Evolution’ in PRACE-2IP is the operation of an infrastructure of national HPC Tier-1 systems in Europe which provides resources to users through PRACE calls for proposals for Tier-1 access (the DECI calls). The Tier-1 infrastructure is tightly integrated with the PRACE Tier-0 infrastructure and for this reason this task is closely collaborating with the corresponding task of PRACE-3IP responsible for the operations of the Tier-0 infrastructure. The services operated by this work package are based on the services listed in the Service Catalogue, which has been defined by WP6 in PRACE-1IP, and which has been updated this year by WP6 of PRACE-3IP. The Service Catalogue is waiting for approval by the AISBL. The Project Management Board has agreed that WP6 should continue to work on the implementation of the services as if the Catalogue had been approved. The management structure of operations is the same as for the Tier-0 infrastructure and the same persons are responsible for particular services on both the Tier-0 and Tier-1 infrastructure. The well established operational procedures and facilities have been successful in the provision of a reliable and highly available infrastructure. In this project period the PRACE Tier-1 infrastructure has increased from 18 to 21 sites. Five new sites were integrated and two Tier-0 sites were removed as Tier-1 resource. Six sites upgraded their systems this project year. The total peak performance of the Tier-1 infrastructure has increased by more than 50% from 3080 to 4865 TFlop/s. The dedicated 10Gbps network is operated by PRACE in close collaboration with GÉANT. Two more sites from Poland have been connected with 10Gbps this year using the existing 10Gbps link to Poznan. Four sites have been connected with 1Gbps to the PRACE network through the newly developed IPSEC service. The network monitoring facilities have been enhanced providing link status information for sites connected through the PRACE network. GridFTP has been deployed and supported at all sites with a link to the PRACE network as the high throughput data management facility for transfers between PRACE sites as well as between PRACE sites and external sites. In addition, some sites support the gtransfer tool to ease the management of GridFTP data transfers. This task supports job submission to the local batch systems and successfully deploys the current version of UNICORE to submit jobs remotely. A common user administration facility for the distribution of information needed for the authentication and authorisation for PRACE services has been maintained and extended with new partners. The helpdesk handled more than 380 Tier-1 related support tickets. The User documentation for PRACE services has been updated with the changes in the infrastructure. The monitoring facilities provided with the Inca tools have been enhanced with several new features and additional facilities, one of which is the information on the version of the deployed services at sites. The second objective is the integration of new Tier-1 sites. Five new sites which started to provide resources for the DECI calls in this project period have been integrated, while for one more site the preparations started but only can be finished later because of a delay in the delivery of their system. This brings the total number of successful integrations of new Tier-1 sites for the whole PRACE-2IP project period to eleven, so almost doubling the total number of Tier-1 sites.