IBM High-Performance Computing and Storage

CCV has installed an IBM parallel computing platform and parallel data storage system. The core system has 166 compute nodes, each with 2  Intel Xeon E5540 (quad-core Nehalem) processors, 24 Gigabytes of RAM, and a QDR (40 Gigabit/s) Infiniband network interface. Infiniband is used both for application message passing and also for disk I/O. Disk access is provided through the IBM General Parallel File System (GPFS), with 12 disk service nodes and 12 IBM DS3400 RAID systems holding roughly 400 Terabytes of disk. A 2 Petabyte Tivoli Storage Manager (TSM) system provides backup and archival storage.  The core computing platform also includes 7 large-memory SMP systems, each with 4 AMD Opteron (quad-core Shanghai series) processors, 64 Gigabytes of RAM, and QDR Infiniband. The system is connected to the Brown campus backbone network via 10 Gigabit/s Ethernet.

 The nodes are diskless, with I/O provided by GPFS accessed through the Infiniband interface. The dual-processor compute nodes are packaged in IBM’s iDataPlex system, which is designed for high reliability and reduced power consumption.

Two dual-processor management nodes provide network provisioning, monitoring and other system management utility. Two login nodes provide access to the cluster for application development, debugging and job submission.

The cluster runs the CentOS build of the RedHat Enterprise Linux (RHEL) operating system, and is provisioned with the xCAT open source cluster management software. Workload management is performed using the Moab Cluster Suite, a commercially supported system for managing queuing and scheduling systems and for user job submission.

The cluster is intended as the core of a new shared HPC and research data storage facility. Investigators are invited to contribute equipment funding under a "condominium model" to enhance both the computing and storage capabilities of the system. In return they will be assigned priority access to that portion of the system which they have contributed. Please ee "Terms of Service" (link to be added) for more information.