Building a Linux-Based High-Performance Compute Cluster

The Rocks clustering package from the University of California at San Diego makes it easy to build and maintain a high-performance compute cluster with off-the-shelf hardware.
Step 15. Install Additional Compute Nodes

Install the rest of your compute nodes. After a couple more nodes are booted, the list of installed nodes looks like Figure 22.

Figure 22. List of Installed Appliances (Three Nodes Added)

When the last node in the cluster reboots at the end of its loading process, press F8 on the head node to finish the installation.

Step 16. Quick Functionality Check

Your cluster now is complete and ready for work. First job: roll call. The Rocks cluster-fork function allows the user to execute the same application on all or a subset of the nodes in the cluster. Figure 23 shows executing the uname command via cluster-fork.

Figure 23. Running uname via cluster-fork

The first invocation requires the system to set up the security for each node. Once this is done, subsequent invocations simply run the application. It appears that all of the nodes in the cluster are healthy and ready for work.

If you are looking for a more comprehensive test, take a look at the Intel Cluster Checker package. This application is useful both on a newly created cluster and as a tool for ongoing maintenance.

Step 17. Install Some Cluster Applications

Now that your cluster is functional, it's time to show it off. One of the more interesting parallel applications is NAMD, a molecular dynamics simulator from the University of Illinois. Paired with VMD, its graphical interface, you essentially have a chemistry set in your cluster.

Summary

When a workstation isn't fast enough, a properly configured cluster can provide all the computing capability you require. Although it is possible to set up a compute cluster manually, many packages are available, both free and commercially supported, that can make the installation and configuration process essentially painless.

Tom Lehmann had a 30+ year career with the Intel Corporation working in just about every facet of the computing industry. The last ten years of his Intel career were spent in high-performance computing, in particular, high-performance compute clustering. He is currently a consultant working on system designs, Linux-Windows cluster integration and a Linux-based mainframe emulator called Hercules. Tom, his wife and two dachshunds live in Las Vegas.

______________________

Geek Guide
The DevOps Toolbox

Tools and Technologies for Scale and Reliability
by Linux Journal Editor Bill Childers

Get your free copy today

Sponsored by IBM

Webcast
8 Signs You're Beyond Cron

Scheduling Crontabs With an Enterprise Scheduler
On Demand
Moderated by Linux Journal Contributor Mike Diehl

Sign up and watch now

Sponsored by Skybot