terabyte ComputingTerabyte ComputingProfessor Martín founded the CRoCCo Laboratory in Princeton University the summer of 2002. The laboratory includeds a unique, state-of-the-art computational facility that is dedicated to the numerical simulation of turbulent hypersonic flows and ranked as the 82nd fastest computer in the world at the time of its installation with a 0.5 Teraflop capability.

The system consists of 128 nodes with each node occupying 1U of rack space. The nodes are dual Intel Pentium Xeon 2.2 GHz processors with 2GB of memory, and are connected by both a Myricom Myrinet fiber switch for parallel communication.

Two data servers are attached to the cluster. The first, a Panasas parallel fileserver, serves 4 Terabytes of data across 10 file I/O servers and a metadata server. Access is across a trunked 4-port Gigabit interface. The compute nodes use a special kernel interface to access these data I/O servers directly thus speeding up access to the data.

The second data server is a traditional NAS storage device supplying NFS service to the compute nodes, again across a 4-port Gigabit trunked interface. This 8 Terabyte filesystem resides across two RAID 6 arrays of SATA-2 disks attached to the NAS host through a 4 Gigabit fiber channel interface.





Back to top      
CRoCCo Lab Home Clark School Home UMD Home