Dell HPCC News

The Top 500 list provides a biannual ranking of the world's most powerful computers. Dell has 27 entries in the June 2008 Top 500 list.

List highlights:

• Dell is ranked 3rd among all hardware vendors in number of list entries.
• Dell has the highest efficiency cluster.
• Although Dell accounts for only 5.4% of the systems on the list, Dell systems account for nearly 10% of the aggregate top 500 performance.

Top 500 HPCC List

HPCC Links

Here are a number of links to help you with your HPCC technical quest:

Dell / HPCC
June 2008 Top 500 List
HPCC Overview - Podcast

Here are some useful links for HPC related news sites, blogs, and otherwise useful URLs:

HPCWire (Best general site for HPC news and information)
InsideHPC (Good site for HPC news)
Clustermonkey (Great site for general cluster articles - Jeff Layton is one of the authors at this site)
Scalability (The best HPC blog on the web from Joe Landman)
HPCCommunity (Good site to talk about LSF, OCS 5.x, and Kusu)
Lead, Follow, or... (Good blog from a gentlemen that has lots of experience in HPC. His blog is updated infrequently, but it's very high quality)

Introduction to HPCC

Some people are very familiar with HPCC but those who aren't or those who want to learn more follow the link below to an Introduction to HPCC. The article starts with a brief introduction to HPCC but also includes links that discuss various aspects of HPCC. Please be sure to follow the links in the article, but come back to this page for more information.

Introduction to HPCC

Power and Cooling

Datacenter power and cooling is another important topic in HPCC solutions, because a
large number of densely populated servers in a data center can generate a lot of heat and consume a lot of power. HPCC solutions tend to be at the extreme end of the range of high density and high performance which makes an efficient power and cooling design a key component.

HPCC Software

Once all of the hardware for a Dell HPCC cluster is purchased, deployed and configured the next obvious step is deciding "What software do I need to run on the cluster?". There are two software components that are always needed:

Applications - ultimately the reason the cluster was purchased in the first place. Typically during deployment and overall cluster qualification, application benchmarks are run to ensure the cluster will meet the customer needs.

Operating System - Red Hat Enterprise Linux or Microsoft Windows CCS. Applications really influence the choice of operating system. If the application runs on Microsoft Windows but not on Linux then the Operating System choice is obvious.

Once the Applications and Operating System are chosen there are many 'other' software components required for an operational cluster all of these components form the Software Stack. Dell HPCC has a nice diagram on the Dell HPCC website that shows a high level HPCC Software Stack.

The Dell HPCC diagram is just an example software stack, other organizations have their own definition of a software stack but in general you need the following layers (in no particular order):

• Basic Cluster Services
• Cluster Provisioning
• Cluster Management
• Cluster Monitoring
• Cluster Events & Logging
• Workload Management
• Cluster Reporting
• Cluster Filesystem and/or Parallel Filesystem
• Patching & Repository Management
• Development tools
• Tools and Libraries for parallel programming
• Application and/or Grid Portals

As you can see that is a lot of components to install, and configure. Fortunately there is movement in the industry to standardize some of the components in the software stack to make it much easier. Intel, Dell and other partners such as Platform Computing, Ansys, CD-Adapco, The Mathworks and others are working together on Intel Cluster Ready.

The Intel site has a good description of Intel Cluster ready:

The Intel® Cluster Ready program and technology package makes it easier to design, build, sell, program, acquire, and deploy clusters built with Intel components. In collaboration with OEMs, channel members, and ISVs, the program specifies a common basis for clusters. This ensures ISV applications written to run on one certified cluster can reliably run on another certified cluster; conversely, a certified cluster will support multiple Intel® Cluster Ready ISV applications.