Clustering with RHEL?

dsschanze · 03-08-2006, 02:37 PM

I am working on a project in tech support at my school and the project will involve setting up a 2-3 32-bit intel processor cluster in order to crack passwords out of very large hash files. My question is, what distribution would this be easiest in? Also, if you could recommend documentation or reading material that would explain how run intel 32-bit clusters, please tell me.

-dss

Wells · 03-09-2006, 08:15 AM

One place to look would be over at ROCKS clusters.

http://www.rocksclusters.org

sylvian14 · 03-12-2006, 11:46 AM

I think Clusters might not be the solution here, could be a Grid solution u're looking for. Cluster is only for hardware redundancy and grid cluster, where you can pool all the resources in all your various PCs/Servers out there to provide the processing power as and when you need it.

dsschanze · 03-12-2006, 06:45 PM

Quote:

Originally Posted by sylvian14

I think Clusters might not be the solution here, could be a Grid solution u're looking for. Cluster is only for hardware redundancy and grid cluster, where you can pool all the resources in all your various PCs/Servers out there to provide the processing power as and when you need it.

so running a boinc server would be a solution then?

Wells · 03-13-2006, 08:44 AM

Quote:

Originally Posted by sylvian14

I think Clusters might not be the solution here, could be a Grid solution u're looking for. Cluster is only for hardware redundancy and grid cluster, where you can pool all the resources in all your various PCs/Servers out there to provide the processing power as and when you need it.

There are a number of different cluster styles. You are right in once sense, in that clusters can be used for high availability. But in another sense they can also be used to work on a problem using multiple processors through the concepts of distributed computing.

The cluster that I work on can be used in a couple of different ways:

Multiple jobs can be created, each one of these jobs performing a task on a set of data. This is your basic batch queueing system. Take a look at OpenPBS for this sort of thing. This can be very powerful for people who have tons of different data sets in which they need to perform a task upon. Each compute node performs the task on a different data set, and less time overall is used to do the process because multiple processors are used.
A single job which has been fashioned in such a way that multiple machines can work on different parts of the job at the same time. MPI and LAM both come to mind for this situation. Basically what happens is you have a program that is written in a special way to be run on multiple machines at the same time. This program passes messages back and forth between the different machines to let each machine know what part of the problem it is working on.

thorn168 · 03-28-2006, 05:45 PM

I would suggest cluster Knoppix in a computer lab with dedicated master node to save the results of the processes running on the "Volunteer" machines. Cluster knoppix works like a live CD distro on "volunteer" machines which do the number crunching work but you need a master node controller to save the work and load balance the jobs.

Check out the Cluster Knoppix link here: http://bofh.be/clusterknoppix/about.htm

You can also google for more information about projects using cluster knoppix.