How to manage the Dutch Tier1 of the LHC

Mark van de Sanden
SARA

Abstract

For the distribution, storing and analyzing of the data from the LHC (Large Hadron Collider) a global collaboration was setup linking large grid infrastructures and computing centers together to build the WLCG (Worldwide LHC Computing grid). The WLCG is a tiered infrastructure with a T0, which is CERN (the Machine), 11 T1s and a large number of T2s. SARA with neighbouring institute Nikhef participate in the WLCG and form the Dutch Tier1 for the LHC.

When running production the LHC generates about 15PB per year. A T1 is next to the T0, the primary site to permanently store, process and analyze about 1/10th of the produced data capacity, which is about 1-1,5PB per year. SARA and Nikhef have been building up the last 5-6 years the infrastructure of the Dutch T1. Currently the shared infrastructure of the T1 consists of a compute cluster about 3000+ cores, online storage of about 4PB, about 1.5 PB offline tape storage, a small cluster running a large number of different grid services built around a 10GE network infrastructure.

In this talk, I will speak about aspects of setting up and maintaining this complex infrastructure, datacenter planning, backups, monitoring of the availabilty and the performance and how to face out components. But I also want to give some highlights in collaborating in a worldwide project.

Biography

Mark van de Sanden started his professional career in 1994 as a system administrator at the National Aerospace Laboratory (NLR). In 1997 he started working at SARA Computing and Networking services managing super computers. Currently he is teamleader of the Mass Storage group, responsible for all mass storage infrastructure at SARA consisting of a multi Petabyte online and offline storage infrastructure.

In 2004 Mark got involved in the WLCG (Worldwide LHC Computing Grid) project, setting up the first generation of the dutch Tier1 storage infrastructure and running the first service challenges. In the last years the infrastructure has evolved to the current large scale computing and storage infrastructure running the Tier1 services.