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.

Um, Ah Cluster Planning

Recently, fellow Cluster Monkey Jeff Layton and I participated in a Cluster Planing pod-cast over at Research Computing and Engineering (RCE-Cast). Jeff and I offer up a few tips that may help you navigate your way through the maze of HPC cluster options and methods. Of course, the word "um" is not to be confused with some kind of cluster feature. Thanks to Brock Palen and Jeff Squyres for putting up with us.

This, That, and HPC Things

A few tidbits that may help brighten your HPC day

A few links that you may find interesting. First, Dr. Dobbs is running an article on the new Larrabee API. If you have not heard of Larrabee, it as a new architecture from Intel aimed at the GPU market. The design involves many x86 cores (Pentium P54C) and vector processing. All cores will be cache coherent. It is a kind of a cross between a multi-core processor and GP-GPU from NVidia or AMD/ATI. Of course HPC guys would never use a GPU for crunching numbers.

While we are talking about number crunching GP-GPU's, NVidia has released CUDA version 2.2 with the CUDA GDB style debugger. That is right, for those rare programmers that might just need a debugger, a CUDA debugger with GDB interface is available with all those features you know an love including breakpoints, watch variables, inspect state, etc., as well as additional functions for CUDA-specific features. The is also an update of the Visual Profiler for the GPU that supports, among other things, full measurement of memory bandwidth within a kernel. There are some other improvements as well.

Finally, our very own Jeff Layton has created a Nehalem Memory Cheat Sheet. Thanks to Dell for helping Jeff create this and thanks to Jeff for providing good clear information at a product launch.

Listen Up: Research Computing and Engineering

If you are interested in HPC Podcasts head on over to Research Computing and Engineering (RCE-Cast). Research Computing and Engineering targets topics relevant to the High Performance Computing (HPC) and Research Computing communities. Some recent topics include Hadoop (free version implementation of MapReduce and Google FS), HOOMD (Highly Optimized Object Oriented Molecular Dynamics), Open-MPI, and others. More to come. Give your Jonas Brothers mp3's a break and listen to RCE-Cast.

Webinar: IDC HPC Predictions 2009

Wondering how it is all going to work out? Here is your chance to hear the leading prognosticators talk about the HPC market. There is a Complimentary IDC Webinar on Wednesday February 18 from 1:00-2:00 p.m. ET.

IDC Abstract: As the global economic recession unfolds, the high-performance computing (HPC) technical server market is feeling the impact. But, how will these economic and market conditions evolve in 2009? Join IDC's expert analysts as they present IDC's top-10 predictions for the worldwide HPC market. This session will examine the trends, challenges, and market opportunities emerging in 2009, and provide essential guidance and advice for finding success in these challenging economic times.

Speakers include: Earl C. Joseph, Ph.D., Program Vice President, High-Performance Computing Steve Conway, Research Vice President, High Performance Computing Jie Wu, Research Manager, Technical Computing Richard Walsh, Research Director, Technical Computing

Learn About Open-MX

From the this is way cool department

Open-MX is a message passing stack that offers the Myrinet Express features on top of regular Ethernet hardware. It requires no kernel or driver modifications. In addition any software that support (e.g. various MPIs,PVFS2) Myricom MX libraries can link to directly to Open-MX.

The project page and software can be found here. Recently, project leader Brice Goglin gave a presentation entitled High-Performance Message Passing over Ethernet with Open-MX, why and how? You can find the slides and video at the Distributed Computing Group (DISCO) site. Quick review: I have used Open-MX since version .4 (current version is 1.04) and it is a well designed and run project that works great If you use Ethernet, check out Open-MX!

Search

Feedburner

Login Form

Share The Bananas


Creative Commons License
©2005-2012 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.