Here’s a collection of highlights, selected totally subjectively, from this week’s HPC news stream as reported at insideHPC.com and HPCwire.

>>10 words and a link

Platform revs LSF;
http://insidehpc.com/2008/05/19/platform-computing-announces-latest-release-lsf-scheduler/

U of Arizona adds 1,392 core ICE with water-cooled doors;
http://insidehpc.com/2008/05/20/university-of-arizona-buys-sgi-altix-ice/

HPC may help feed hungry with better rice;
http://insidehpc.com/2008/05/20/feeding-the-hungry-with-better-rice/

Altair’s HyperWorks platform opened to ISVs;
http://insidehpc.com/2008/05/21/hyperworks-platform-opened-to-third-party-developers/

Inaugural CUDA coding contest offers $5,000 cash;
http://insidehpc.com/2008/05/21/cuda-coding-contest/

NEC to upgrade Earth Simulator;
http://insidehpc.com/2008/05/18/earth-simulator-to-be-upgraded-to-131tf/

>>Microsoft Announces Windows HPC Server 2008 Beta 2

Microsoft has announced that it will soon begin shipping the Beta 2 of its Windows High Performance Computing Server 2008 product. Beta 2 was a big feature push for Microsoft as it strives to compete with several, well-vetted Linux competitors.

“With Beta 2, we provide a highly available head node that integrates deployment, management, monitoring and diagnostics in a new user interface based on System Center’s ‘Microsoft’s systems management tools’ user interface framework,” wrote Ryan Waite, group program manager for Microsoft HPC, in a blog post on the Windows Server Division Weblog.

Since the initial Beta 1 launch of WinHPC2008, Microsoft has done several tests on clusters of up to one thousand nodes and fixed more than one thousand bugs. So far, they are on track for a general release in the second half of 2008.

Read the full article here.

>>IBM buys into the HPC interface

Some of you may have been forced to sit through one of the religious sermons I’ve delivered over the past several years proselytizing the idea that HPC has a really big problem: our interface sucks. I mean, a command line. Really?

The problem is not that the command line is inherently hard. The problem is that it is in the way for generations of new researchers who are going to replace all you crusty bastards who are saying to yourselves “anyone who wants to do HPC and doesn’t want to deal with shell scripts and command lines can’t handle an account.” Anything in the way of a million more people using HPC has to go.

Anywho, for the two of you still reading, here’s the story:

IBM said on Wednesday it has started selling software that lets customers access its Cognos business intelligence software via BlackBerry mobile devices.
…

The Cognos program, which sells at a list price of $300 per user, allows customers to view real-time analytics on the state of their business on their BlackBerrys.

There you go. Big enterprise iron tied to ubiquitous mobile platforms. I predict you’ll be seeing much, much more of this type of thing.

>>LS-DYNA teams with Evergrid for transparent checkpointing

Evergrid and Livermore Software Technology (makers of LS-DYNA) announced this week they’ve teamed up to integrate Evergrid’s Availability Services into LS-DYNA:

With the combination of Evergrid AvS and LS-DYNA, customers can checkpoint a running job at any time during execution without stopping or pausing the job. The job can also be preempted for a higher priority project, freeing up system resources, such as CPU, memory and the LS-DYNA license. After a system failure or completion of the higher priority project, the job can be restarted seamlessly from the last checkpoint without wasting valuable compute cycles. “With Evergrid AvS, LS-DYNA customers can ensure that they can meet their projects’ deadlines even with unexpected system failures or other interruptions,” said Dr. Wayne Mindle of LSTC.

After several years, it looks like Evergrid’s technology may finally be on its way out of development, which is a good thing. They have much bigger plans for their software than simply restoring a single custom outfitted-app — if they get it all developed according to their original plans, users won’t  have to worry about modifying their codes for this kind of support, and administrators will be able to stop, preempt, and move apps around in real time without losing state.

—–

John West is part of the team that summarizes the headlines in HPC news every day at insideHPC.com. You can contact him at [email protected].