Penguin gets personal with Linux clusters

Heavy beast slides under your desk

By Ashlee Vance in Seattle • Get more from this author

Posted in Servers, 17th November 2005 23:16 GMT

Free whitepaper – Avoiding costs from oversizing data center and network room infrastructure

SC05 Penguin Computing has joined the deskside cluster movement with a new box that packs tons of horsepower in a package meant to be convenient for engineers.

The Penguin Portable Cluster stacks numerous blade servers in a vertical, tower-like configuration. Customers can fill the system with anywhere between 6 and 24 low voltage Xeon or Opteron processors – single or dual-core on the AMD front. The system also supports up to 96GB of memory, holds 25 disk drives and has a non-blocking Gigabit Ethernet fabric.

The impressive Scyld cluster management software ships with the system.

The Portable Cluster isn't too different from the deskside cluster sold by Orion Multisystems. It is, however, less of an integrated package as it doesn't have a single "power on" button or tightly-linked internal components. The Penguin system also packs some serious weight. We put our full muscle into lifting the system and could only get it two inches of the desk. Our feebleness aside, you'll need a forklift or California's governor to shove this thing under a desk.

Engineers love these new types of systems as they provide a personal-sized cluster. They can run smaller jobs at their desk or in a lab and transfer work with relative ease to larger clusters designed with similar architectures such as Penguin's Performance Cluster line.

Microsoft chief Bill Gates put his good name behind the personal cluster idea during a keynote earlier this week here at the Supercomputing conference.

You can check out the Penguin system here, and we recommend that you do. It's a cool looking box.

Penguin told us that about 80 per cent of its sales come from Opteron-based systems. "No one is asking for Intel in these types of systems," a rep said. ®

Free whitepaper – Standardization and Modularity in Network-Critical Physical Infrastructure