Dear Desktop Engineering Reader:
Clusters have changed the nature of engineering in small and mid-sized outfits (and lots of big ones) by the way they band together high-performance computers to create supercomputing power. But clusters are notoriously finicky contraptions, with the most common user complaints being that they're never available. The IT guys will tell you that they have compatibility troubles and that managing a cluster of clusters is a wrestle. With its introduction of its new Xtreme-X2 Supercomputer, Appro is out to make clusters reliable, easier to manage, and more efficient. That is, Appro wants to make clusters operate like you know they should.
The 64-bit dual-socket, Quad-Core AMD Opteron processor-based Xtreme-X2 is built with a scalable, flexible architecture that enables you to group together clusters to make what Appro calls a "Scalable Unit." You can then manage this integrated Scalable Unit of clusters as if it were a standalone supercomputer. Your key to managing all that is Appro's Cluster Engine software, which comes pre-installed with every Xtreme-X2 system. This software handles administration of the entire system, including the network, server, clusters, and storage subsystems. Cluster Engine software uses a web-based interface that lets you run, check, even reboot the whole system remotely.
The Xtreme-X2's specs are impressive: up to 128 nodes/512 Opteron processors and 4.5-teraflops (TF) of computing power in a single 44U equipment rack. It has 4TB of memory, hot-swappable blades, redundant cooling fans, redundant power supplies, and Infiniband connections with low latency Mellanox ConnectX host channel adapters. Appro also says that the Xtreme-X2's new “Directed Airflow” cooling configuration reduces datacenter floor space by 30 percent while maximizing power and cooling efficiency.
I'm not alone thinking that the Xtreme-X2 is impressive. ING Renault Formula 1 (F1) Team’s Computational Fluid Dynamics (CFD) Centre in the UK has already inked a deal with Appro for a 38TF configuration to run full-car simulations and critical-components aerodynamic tests. You can read about that and get all the specs on the Xtreme-X2 Supercomputer from links in today's Pick of the week write-up. You'll also find some links to multimedia presentations on the Scalable Unit concept and cluster management from the Supercomputing 2007 conference. While there, you can sign up for an IDC white paper on the Xtreme-X2 as well.
Appro describes its Xtreme-X2 as built for HPC. The more I learn about the Xtreme-X2, I'd add that Appro is also making every effort to make the promise of high-performance clustering what everybody always said it should be and maybe more.
Thanks, Pal. -- Lockwood
Anthony J. Lockwood
Editor at Large, Desktop Engineering Magazine