Is it possible to cluster say five or six machines, of all different speeds and types, then F@H over them all. I have a strange feeling my two P2 machines and three P3 machines could possible run as fast as maybe one of my semprons thisway, thusincreasing productivity. (since the move to 50k steps and such on WU'sthey have a month to finish one WU)

S1ugh34d I've considered it many times but most people I've spoken to shot down my idea but hopefully somebody will come up with a better answer in the meantime when I was considering doing a cluster I found a pretty cool site

ClusterMonkey

Nice link, i found a link at there site to PK. Paralell Knoppix, the guy in this tutorial ran GROMACS, which is what F@H uses to fold (or crunch)

Although im a bit skeptical about the whole setup, being that F@H normally uses one core, thus why dual cores are unaffected by one instance running. but ill give it a swirl, maybe ill get these PC's o mine running a nice cluster, folding at will.

PC's hopeflly going into the setup:

P2 366MHz @ 396MHz
P3 C 900MHz
P3 866MHz
P2 350MHz
Sempron 3100+ @ 2.5GHz (master node)

I'l post results in this thread, being that the 3100+ hasnt seen a newer WU (ie 50k, 200k steps) i couldn't say how fast it used to run, but I'm sure if it works, itll run faster.

Will this link help?

+Rep for you The Captain ! I have to look into this in more detail now

Quote: Originally Posted by s1ugh34d Is it possible to cluster say five or six machines, of all different speeds and types, then F@H over them all. I have a strange feeling my two P2 machines and three P3 machines could possible run as fast as maybe one of my semprons thisway, thusincreasing productivity. (since the move to 50k steps and such on WU'sthey have a month to finish one WU)

I'm catchin up to ya!!

Nice linkage, i actually got PK running nice and have a nice little setup going, one issue with thatfarm setup guide, you have to have a dedicated switch + the main node needs to have a internetconnection that the other nodes don't share. Another thing is it doesnt literally span the work over the different PC's, it just gathers all the work each does and send it from a centralized server. Kinda like a local network that only does one thing.

Quote: Originally Posted by corruptjoker I'm catchin up to ya!!

yes i see that, you'll soon pass me by
although i just added a 2600M barton to the mix, shame my X2 is down, or i might have a chance to keep the lead

This question comes up every once in a while. The best solution I have heard of so far is a Beowulf cluster - I know that HTRN was looking at building one a long time ago.

Quote: Originally Posted by PurpleDiamond This question comes up every once in a while. The best solution I have heard of so far is a Beowulf cluster - I know that HTRN was looking at building one a long time ago.

Beowulf HOWTO

Beowulf HOWTO

Quote: Originally Posted by PurpleDiamond This question comes up every once in a while. The best solution I have heard of so far is a Beowulf cluster - I know that HTRN was looking at building one a long time ago.

Yep thats what the clustermonkey link project vegas linked me at first, a guide to beowulf clustering. In turn that site leaded me to PK (parallel Knoppix) and it got me4 CD's that are running my cluster diskless, but i still havent tweaked the SMP client over the cluster, working on it though

(don't have alot of time anymore school+work+tring to do things outside of online)

Quote: Originally Posted by THRASHER2 Beowulf HOWTO Beowulf HOWTO

Thanks for the linkage