Machines

The PRObE project will provide two classes of computing resources: large research clusters (older generation technology repurposed from the DOE's Los Alamos National Laboratory) and newer, more advanced hardware (newly purchased). Researchers may obtain dedicated use of as much as a complete cluster for their research.

Here is a summary of the machines acquired, being acquired or proposed over the next few years. While this list is not final, it provides an example of the kinds of machines in the pipeline for use by researchers through the PRObE facility.

• Winter 2011/2012: Kodiak (2048 cores)
• 1024 Nodes, Dual Socket, Single Core AMD Opteron
• 4 GB RAM per core
• Fat-tree SDR Infiniband high-speed interconnect
• 1 Gigabit Ethernet
• Winter 2011/2012: Denali (256 cores)
• 128 Nodes - Kodiak front-end cluster
• See above for details
• Summer 2012: Sitka (2048 cores)
• 1024 Nodes, Dual Socket, Single Core AMD Opteron
• At least 4 GB RAM per core
• Fat-tree Myrinet high-speed interconnect
• 1 Gigabit Ethernet
• Spring 2012: Yukon (256 cores)
• 128 Nodes - Sitka front-end cluster
• See above for details
• Fall 2013: Nome (1600 cores)
• 200 Node, Quad Socket, Dual Core AMD Opteron
• 2 GB RAM per core
• Fat-tree DDR Infiniband high-speed interconnect

In addition to these high node count clusters, PRObE envisions purchasing a series of new high core count machines. All hardware is estimated and will be modified depending on the actual machine acquired.