Clustercorp Brings Folding@Home to the Cloud for SC10

[ParrLeyne]

Came across this in an RSS feed I follow:

"Clustercorp and its partners have joined to provide free EC2 instances to run Folding@Home in the cloud with Rocks+ during Supercomputing 2010 (SC10)..."
http://www.hpcwire.com/topic/systems/Cl ... 85929.html

[toTOW]

Too bad they don't allow us to configure the clients running in the instances, I looked at their stats, and they're only running one SMP WU under the username you provide and their team ... no bonus to expect in these conditions (but I don't know if they're even aware of the bonus ...)

[bruce]

Are they aware of the Final Deadline?

There's no doubt that the free processing will be lower priority than everything that they'll be able to sell to a paying customer so the speed isn't going to be something to write home about but at least some of the completed WUs will be a benefit to FAH.

The Queue of work that hasn't started yet seems to be growing faster than the list of WUs that have been completed. I guess that's good.

[nitrobass24]

So are they actually running F@H in a cluster configuration? or do they just have a bunch of linux boxes running F@H?

[Tobit]

So are they actually running F@H in a cluster configuration? or do they just have a bunch of linux boxes running F@H?

It is nothing more than EC2 instances like you can buy via Amazon.com.. it is not a cluster setup at all. Additionally, this is only occurring while the conference is running so it's pretty useless.

[MJandDave]

So, dumb question. Can an individual sign up for the free Amazon Web Services, and run F@H under their Free Usage Tier? If so, how would you do that?

http://aws.amazon.com/free/

[tim.mcintire]

Hi All -- the demo for SC10 was managed as a Rocks cluster (i.e. one frontend node was used to coordinate the provisioning of compute nodes and the job queue). When a username was submitted, a new compute node was created on the fly and provisioned with Rocks+ -- the work unit was then delivered to that individual node for processing. Upon completion, the node was terminated, and a new node was provisioned to run the next job in the queue (part of our goal was to demonstrate the automatic provisioning of a cluster in EC2 using Rocks+).

If you're interested in setting up Folding@Home to run on an individual instance, I'd recommend just starting with a reference Amazon Linux AMI and installing/launching FAH by hand -- very simple if you are just running one instance. The latest version of FAH will run on CC1 instances, but on the less expensive Micros, I'd recommend FAH504-Linux.exe (note: these take a very long time to complete, as these instances aren't designed for heavy processing).

-Tim

[MJandDave]

Thanks for the info. What do you mean by a "very long time to complete"? Would a classic client be completed by its deadline?