Select News

The news in this category has been selected by us because we thought it would be interestingto hard core cluster geeks. Of course, you don't have to be a cluster geek to read the news stories.

In case you missed it, Introduction to Beowulf Design, Planning, Building and Administering running again in November (7th-10th). As before the course is presented by the Advanced Research Computing (ARC) team at Georgetown University. The first session in September was success and the we can look forward to other trainings as well.

Quads are old news, 80 cores are what you really want

In addition to all the other hoopla at this weeks Intel Developers Forum (IDF) there was the announcement of Intel's Tera-scale Computing effort. Other than being on of the more over used HPC buzzwords, it does propose some interesting directions beyond, the current core race. As their white paper points out, you can't just keep slapping cores on the processor die. Interestingly, the 80 core processor design arranges cores in a grid and provides communication to four neighbors. Flashback to the Inmos Transputer.

All the cool hardware aside, the real issue is software. As soon as two or more cores/processors/servers etc get involved, there is a need to create parallel software. If you compared parallel software technology to today's hardware you would be looking a 80286 (maybe a 386 at best) processor. (And I'm not even going to get into scaling issues). A multi-core detente of sorts, as proposed by myself (see the October issue of Linux Magazine and Tom Yager at InfoWorld may be A Good Thing (tm) to consider.

The core wars, begun they have

This week at the Intel Developers Forum (IDF) several vendors are showing the latest in multi-core fashion -- the quad-core Xeon. Dubbed the Clovertown, this quad-core beastie is mentioned as part of Intel's new Core™ Microarchitecture". In particular, Appro is demonstrating a 1U server with two quad-core Clovertown processors. That is a total of eight cores in a standard rack-mount enclosure. I happen to know that they also have two of these systems connected via a Mellanox infiniband HCA and they are running the NAS Parallel Benchmark Suite under Linux on all 16 cores. Rumor also has it that the monkey who set up the demo is running GKrellM on both nodes and 16 little windows show busy little cores. "What about numbers?, you ask. Stay tuned, real numbers for real cluster applications will be available in November at SC06

And, by the way, hypothetically speaking, if a certain Cluster Monkey happened to have these two nodes for a few more weeks, what would you like to see him run on it? (add your suggestion as a comment) That is, of course, hypothetically speaking.

They named it Roadrunner, wonder if they include any Acme earthquake pills or rocket sleds. Just a thought.

IBM has announced the "worlds fastest supercomputer" that is to be built from Cell processors. Normally, I don't post or comment on fastest supercomputer headlines, but this announcement represents an interesting take on the HPC market. First, I think they forgot to tell us what program(s) this machine will run the fastest. My guess is it will be HPL and a trip to Top500 glory, but I could be wrong and that is not what really matters here.

There is an interesting discussion about Parallel Languages over at the IEEE Technical Committee on Scalable Computing. The discussion was initiated by Greg Pfister of IBM (And author of In Search of Clusters.) The excellent question he posed is below (click Read More) Your thoughts are welcome here as well.


Login And Newsletter

Create an account to access exclusive content, comment on articles, and receive our newsletters.


Creative Commons License
©2005-2019 Copyright Seagrove LLC, Some rights reserved. Except where otherwise noted, this site is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 2.5 License. The Cluster Monkey Logo and Monkey Character are Trademarks of Seagrove LLC.