CPlant - FY 98 Platform

Phase I Platform

During 1998, Sandia purchased an 25 scalable units from Digital Equipment Corporation (now Compaq). DEC used the Cplant 97 prototype scalable unit as the basis for constructing and integrating the new hardware.

Software Components

Compute

All compute nodes currently run Red Hat Linux 5.1 and version 2.0.34 version of the Alpha/Linux kernel.

High-Speed Network

The Myrinet Control Program (MCP) and driver were developed by Sandia to support Portals.

System Diagnostic and Support

The diagnostic and support nodes have shell scripts that logically partition the compute nodes so that each system software developer can have their own versions of kernels and modules.

User Support

The Cplant compilers are simple shell scripts that call the Digtal UNIX compilers with the appropriate paths to locate the support libraries and MPI library.

Hardware Components

Compute

The compute partition is composed of 400 Digital Personal Workstation 500as, each of which contains:

The workstations do not have floppy drives, hard disk drives, or CD-ROM drives.

High-Speed Network

The 400 compute node machines are connected with Myricom's Myrinet gigabit networking hardware. Each node contains a SAN PCI card connected to a 16-port SAN/LAN switch . The topology of the network is roughly a cube of hypercubes.

Secondary Storage

The I/O partiton is composed of 6 Digital AlphaServer 1200 s, each of which contains:

These were chosen primarily because the dual PCI bus allows for full bandwidth for both the Myrinet PCI card and the SCSI controller.

System Diagnostic and Support

The system support network is composed of 27 Digital Personal Workstation 500a 's, each of which contains:

These nodes are identical to compute nodes, except that each has a hard disk drive.

There are 54 terminal servers that provide console access to all of the compute nodes and level 0 system support stations.

There are 54 power control units that provide the ability to power cycle all of the compute nodes and level 0 system support stations.

There are 54 10/100 BaseT hubs that provide the diagnostic network for managing the machine.

Integration

The machine is composed of 54 cabinets, each of which contains the following:

In addition, cabinets containing compute nodes also contain:

User Support

The compile partition is composed of 1 Digital AlphaServer 4100 that contains:

This Digital UNIX machine is the file server for the cluster and also serves as the compile server for building the executables that run on the cluster.