Computer Cluster

The SUNCAT cluster began operation at the end of September 2010. It is hosted at the SLAC computer center, which made important contributions to its smooth deployment. The computing facilities are located in the computer center building at SLAC (Building 50).There are 284 compute nodes with 2.67GHz Intel Nehalem X5550 processors and 24GB of memory, as well as 64 nodes with 2.67GHZ Intel Westmere X5650 processors and 48GB of memory, for a total of 3040 computing cores. A variety of electronic structure codes are installed: GPAW, DACAPO, Quantum Espresso, VASP, FEFF, AIMS, and FDMNES.The network interconnects are standard Gbit ethernet. Calculation results are stored on an NFS fileserver with 15TB of space, formatted with a RAID6 filesystem that can tolerate up to 2 simultaneous disk failures without compromising data integrity. The data are also backed up daily to two independent sets of tapes. To gain access to the cluster there are 2 login nodes, which are very similar to the compute nodes. This is where users submit batch jobs and analyze results. The farm runs packages (described in more detail here) such as GPAW,DACAPO,JACAPO as well as various support programs like VMD. These programs are parallel applications that use openMPI and they typically run on 16-64 cores for a period of about 2 days. The batch system is the standard SLAC LSF.

In addition to the local resources, we are beginning to use Amazon Web Services (AWS) for first-principles computer simulations.