I remember reading about Linux connecting 60 PCs to essentially work as one unit. I was wondering about the posibility of this being done with say, three PCs to work on something like Folding? Apologies but I genuinely have little idea about this and was hoping someone could point me in the right direction RE: advice/reading material. Many thanks Zinf.
Sounds like your generally talking about farming, such as rendering farms. the folding client is already a distributed app, due to it sending work unit's out to clients. If you wanted it for folding, you'd be better off running seperate folding clients and let them rattle through their own work units, as the client is already "farmed" out from the folding servers. Generally, From my lacking understanding of farming computers, they're normally made up of a controller and then multiple slave's they assign work to, the controller co-ordinates the work and re-integrates it together. have a search out for linux rendering farms and see what it turns up.
Correct. Just found an article which looks like a pretty good intro to Linux Clusters: http://www.squidoo.com/linux-clustering. Of course, there's always google if you need to know more However, I agree with WarMadMax in that I'm not really sure how much benefit it would give you for F@H. As he's mentioned, F@H is already a distributed client - effectively, every machine running the client is part of one exceedingly huge cluster. Where clusters come into their own is with things like Beowulf clusters; using a network of relatively inexpensive PCs to run massive computational tasks. You can (with the right hardware) get supercomputer-like computational ability out of relatively cheap hardware - relatively cheap when compared to the cost of a supercomputer, that is.
You can't fold on a cluster (trust me, I've looked in to it). There's not enough speed between nodes to enable the client to work properly. Rendering on the other hand works well.
Wrong. a) Folding doesn't support clusters b) Enterprise supports load balancing & failover. You need hpc server for this type of clustering.
