The Supercomputing Institute currently provides supercomputing resource allocations and technical support for Calhoun (SGI Altix XE 1300 Linux Cluster), the IBM BladeCenter Linux Cluster, SGI Altix, IBM Power4, and IBM Netfinity Linux Cluster.

The Institute has purchased a SGI Altix XE 1300 with 256 compute nodes. Each node has two quad-core 2.66 GHz Intel Clovertown processors sharing 16 GB of memory. This gives a total of 2048 cores on the system and an aggregate of 4.1 TB of RAM. The system is being installed and will be in production in July.

The Institute has purchased a Linux Cluster from IBM. It is a IBM BladeCenter H* with 309 LS 21 nodes. Each node has two dual-core 2.6 GHz AMD Opteron processors sharing 8 GB of memory. This gives a total of 1236 cores on the system. This system is now in production. There is approximately 40 TB of disk on this system. The Institute plans to emphasize parallel jobs on this new Linux Cluster.

The SGI Altix Cluster consists of 11 shared memory machines:

• One 256-processor compute server with 512 GB of memory and 1.6 GHz Intel Itanium 2 processors

• One 48-processor compute server with 96 GB of memory and 1.3 GHz Intel Itanium 2 processors

• One 4-processor interactive machine with 12 GB of memory and 900 MHz Intel Itanium 2 processors

• 8 16-processor Altix 350 compute nodes with 32 GB of memory and 1.5 GHz Intel Itanium 2 processors

All machines run the Linux operating system and share 4.5 TB of disk space. An additional 4.5 TB of disk space is dedicated to local scratch spaces on the nodes.

The IBM Power4 system is a constellation of several shared-memory nodes. Users submit to a single queue on an interactive node, and a fair-share scheduling system schedules each job on an appropriate available node, or users may request a particular node or nodes (see http://www.msi.umn.edu/power4/quickstart.html for details). The operating system is AIX, IBM's brand of Unix. The nodes are either pSeries 690+ nodes (called Regatta nodes) or pSeries 655+ nodes, and the detailed characteristics of the nodes are as follows:

• One 32-processor p690+ node with 128 GB of memory and 1.7 GHz Power4 processors

• One 32-processor p690+ node with 64 GB of memory and 1.7 GHz Power4 processors

• One 32-processor p690+ node with 64 GB of memory and 1.3 GHz Power4 processors

• One 24-processor p690+ node with 24 GB of memory and 1.3 GHz Power4 processors

• Thirteen 8-processor p655+ nodes with 16 GB of memory each and 1.5 GHz Power4 processors

• Eleven 8-processor p655+ nodes with 16 GB of memory each and 1.7 GHz Power4 processors

The interactive node is one of the 8-processor p655+ 1.7 GHz nodes and there are four 4-processor p655+ nodes used as file servers. All nodes are connected to an IBM High Performance Switch (HPS) network.

In summary, this machine has 32 nodes, 344 processors, and 680 GB of memory. There is also 17.3 TB of disk space shared across all 32 nodes.

The Institute's Linux Cluster consists of 76 2-processor Netfinity nodes from IBM. Characteristics of the nodes are:

• Twelve 2-processor nodes with 1.5 GB of memory and 2.6 GHz Intel Pentium 4 processors (Gigabit Ethernet).

• Sixty-two 2-processor nodes with 3.25 GB of memory and 1.26 GHz Intel Pentium III processors (Fast Ethernet and Myrinet).

• One 2-processor node with 2 GB of memory and 1.13 GHz Intel Pentium III processors (Fast Ethernet and Myrinet) (interactive node).

• One 2-processor node with 2.6 GB of memory and 2.66 GHz Xeon processors (Fast Ethernet and Myrinet) (server node).

Sixty-nine nodes are available for computation through a queuing system, three nodes are for interactive use, and the remaining four nodes are file servers. The nodes are connected using Gigabit Ethernet, Fast Ethernet, and Myrinet, as noted above. A cluster file system provides 2.5 TB of disk that is shared by all nodes.

The IBM SP was retired on March 13, 2006.