BEOWULF COMPLETE NEED HELP RUNNING DC!

[paralazarguer]

WTF is up with the servers? We can't even see our stats. I'm afraid to make a dump....

 

[Baldy18]

That is an awesome setup. I wish I could get ahold of some pc's that cheap.

 

[paralazarguer]

NO hard drive and no video card. Would take some money to make them full PCs.

 

[Kris]

Props FRed. That is a significant accomplishment.

How efficient is it?

 

[ProviaFan]

Originally posted by: n0cmonkey

Originally posted by: FlorianRed

just curious, but how are you running eccp on all of those? Are you using a separate client on each one, or what?

No...that would just be a bunch of computers networked together. I thought it was already kind of explained....Um, scroll up and click on FAQ. It describes in minor detail the basic operation of a beowulf.

Did you modify the client at all for this?

Exactly what I was curious to know about, except perhaps I wasn't clear enough. And yes, I do know basically how a beowulf cluster works. It seems that the client would have to be rewritten to take advantage of such a system, since (afaik) the linux client for ECCp is not multi-threaded, and a multi-threaded client would be required to take advantage of more than one system in a beowulf cluster, or else you'd need to be running multiple clients (which could perhaps all be started on one machine, but migrate from there across the network to various machines depending upon their workload - at least that's the way it works with Mosix).

 

[paralazarguer]

Props FRed. That is a significant accomplishment.

How efficient is it?

Actually, it's not that efficient at all. Not at all. It's producing about as much as two athlon XP 1700+ do in 24 hours. I just dumped more than 500points from a 24 hour period. Not too bad I guess.
jliechty, you're on to the right idea. You can run as many copies of the application on the control node as you want unlike in windows. You just have to specify a different file space for each copy (and pray never to reboot heh heh) and all of the proccessing will happen on that machine. Unfortunately, it writes off to a floppy but all you have to do is copy the dplist to control node and submit for each. It's just a matter of mounting all the remove floppies and using their filespace for each copy. Technically, it's not multithreaded but it is centrally controlled so I guess it's beowulf...whatever, it's fast anyway 😀

 

[paralazarguer]

wait a minute....two athlon XPs can't do over 500 in 24 hours! It's pretty fast anyway...

 

[paralazarguer]

Rank within team 98
User FlorianRed
DP 2302
New DP 541
Days inactive 0
😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀😀

 

[MereMortal]

Which version of the client are you running? I would have expected 650-675 DP/day.

 

[Migroo]

Nice one!!

 

[ProviaFan]

Looking good. 😀

 

[joinT]

Originally posted by: FlorianRed
We also have unlimited free healthcare...🙂
Best....country....ever!

😀

 

[n0cmonkey]

Originally posted by: FlorianRed

Props FRed. That is a significant accomplishment.

How efficient is it?

Actually, it's not that efficient at all. Not at all. It's producing about as much as two athlon XP 1700+ do in 24 hours. I just dumped more than 500points from a 24 hour period. Not too bad I guess.
jliechty, you're on to the right idea. You can run as many copies of the application on the control node as you want unlike in windows. You just have to specify a different file space for each copy (and pray never to reboot heh heh) and all of the proccessing will happen on that machine. Unfortunately, it writes off to a floppy but all you have to do is copy the dplist to control node and submit for each. It's just a matter of mounting all the remove floppies and using their filespace for each copy. Technically, it's not multithreaded but it is centrally controlled so I guess it's beowulf...whatever, it's fast anyway 😀

Why not just boot them off floppies or the network (like you seem to be doing) and have them mount NFS? All the points would be on the one system and the floppies wouldnt need to be touched. Run RedHat 7.3 (8.0 beta?) and you may even get to use MereMortal's gcc 3.2 build of the eccp109 client. It wouldnt be a beowolf, but a crackrack is a crackrack 😀