SGI's new super for the enterprise, getting out of the multicore software crisis, IBM's Cluster 1350 upgrades, and more

In today's enterprise HPC news roundup we've got quick links to the major big iron announcements coming out of ISC in Dresden, help getting out of the multicore software crisis, SGI new super for the enterprise, and much more.

Here’s a collection of highlights, selected totally subjectively, from the recent enterprise HPC news stream as reported at

There is a lot of activity at the top end of HPC coming out the International Supercomputing Conference in Dresden, Germany, this week. It’s big news, but not necessarily big news in enterprise HPC, so here are some quick links to the biggest announcements from the conference. Click through for details if you’re interested.

  • Sun announced its PFLOPS-capable Constellation system along with the ginormous 3,456 port switch that makes it all hum. More on this HPC news item
  • IBM announced Blue Gene/P, a rev to its Blue Gene/L high end solution that will scale up to 3 PFLOPS. More on this HPC news item
  • Scali MPI Connect will be available with Windows CCS 2003 on July 31; the new version will include cross-platform compatibility. More on this HPC news item

Getting us out of multicore software crisis

Intel announced a new arrow in the quiver of things it’s doing to avert the coming multi-core software crunch.

Intel Corporation unveiled today a global training curriculum for software developers to speed and expand software development for multi-core processors. The Intel® Software Training Program offers classroom courses and hands-on lab work to better train software developers in multi-threaded application design.

Details on the effort as well as pointers to the course catalog in the company’s press release.

Meanwhile, Savas Parastatidis has a post on his blog pointing to the presentations and papers from the just-concluded Manycore Computing Workshop 07, sponsored by his team at Microsoft.

IBM to announce Cluster 1350 Opteron upgrades

Ashlee Vance over at The Register ran a piece earlier this week with advance news that IBM is upgrading its Cluster 1350 line.

…customers will find the 1U x3455 available with 95W dual-core Opterons running at 1.8GHz, 2.6Ghz and 3.0GHz. That’s because IBM has picked up the fresh Opterons AMD released in April that kicked the dual-core part to 3.0GHz for the first time. So, a similar situations holds true for the 4U Opteron-based x3755 system.

Complementing the Opteron systems, IBM ships the Cluster 1350 unit with Xeon-, Power- and Cell-based servers too.

IBM has also added new networking options with IB from Cisco and Voltaire.

SGI’s big announcement is big news for the enterprise

SGI also launched a new product yesterday (in DC, not Dresden), that very much matters for large computing in the enterprise.

ICE is SGI’s second generation blade solution. It’s Xeon-based (set for dual- or quad-cores) and is configured more like a traditional cluster than SGI’s previous offerings; i.e., no shared memory on this one. Where the 4700 is Itanium-based and will live on to aim at the very high end of the market, ICE is aimed at enterprises and departmental solutions where power, cooling, and manageability are drivers.

The system looks really interesting. You can get 512 cores in 128 nodes (on quad-core Xeon chips) in a single rack. A rack is cable-less up to 128 nodes (all the connections are by a passive backplane), which should smooth installation and management.

Nodes are interconnected by on-board DDR InfiniBand configured as a torus; 4x at node, and the switch blades can be configured with either 4x or 12x. Putting the InfiniBand on board should improve reliability, eliminate concerns about PCI saturation, and also saves on power and space in the configuration.

The energy picture is also interesting; one fully populated rack will draw about 45W/core, or about 23KW. The 4700 draws about 65W/core, so this is moving in the right direction.

You can find out more on SGI’s product page and in the press release.

HP and Microsoft tighten partnership

Today HP and Microsoft are announcing that they’re expanding their relationship. HP is also adding some installation glue and documentation to wrap the whole package up, and both companies are investing in technology centers around the world

As part of the alliance, HP and Microsoft have established technology centers in Houston and Grenoble, France, for customers and independent software vendors. The technology centers provide a venue for benchmarking, testing and validating applications on HP servers based on AMD and Intel processors running Windows CCS.

More on this enterprise HPC news item.

TotalView rev adds support needed for enterprise HPC

HPCwire let us know that TotalView has revved their eponymous debugger this week to version 8.2.

TotalView is a source code and memory debugger that helps developers get at problems in data-intensive, multi-process, multi-threaded, or network-distributed applications. Highlights of the new features include support for

  • SiCortex
  • Cray XT4
  • Fedora Core 6
  • Expanded Mac products
  • Ubuntu

The company also announced that it is working with IBM to bring TotalView to IBM’s new Cell-based QS20 BladeCenter Cell Broadband Engine platform, something sure to be of interest in the financial sector.

Cluster Resources offers Escalante to make cluster deployment point-and-click simple

Cluster Resources, maker of the Moab Cluster Suite, announced yesterday [PDF] from Dresden that their new HPC software stack is soon to enter an open beta.

Escalante leverages SUSE Linux Enterprise Server’s pattern deployment capability to apply the needed software to the cluster’s central management node (a.k.a. head node), then upon restart a Moab cluster deployment wizard queries the user about the number of compute nodes in the cluster, how many nodes there will be per rack and the IP address. While additional configuration and flexibility is available in an advanced section, the above questions are all that are needed by Escalante for a default cluster deployment. At this point, the wizard automates the software configuration and all that remains is for the installer to power on the compute nodes in a logical order.

Cluster Resources is also aiming at some of the low-hanging HPC pain points that small organizations feel when moving up from PCs

From a usability perspective, this product also provides the technologies that make it mainstream-ready. Escalante incorporates Moab Cluster Suite®, which includes an easy-to-use Web-based job submission portal, a graphical Windows-like administration interface and professional reporting capabilities. The professional reporting provides managers with the capacity planning charts as well as resource usage and sharing reports that keep them on top of what is really happening with their investment and what service levels are being provided to their users, projects and associated groups.

John West summarizes the HPC news headlines every day at, and writes on leadership and career issues for technology professionals at InfoWorld. You can contact him at