Welcome to MilkyWay@home

How are you guys able to run so many diff project at once effectively?

Message boards : Number crunching : How are you guys able to run so many diff project at once effectively?
Message board moderation

To post messages, you must log in.

AuthorMessage
Chris
Avatar

Send message
Joined: 3 Oct 10
Posts: 42
Credit: 320,242
RAC: 0
Message 47148 - Posted: 7 Apr 2011, 18:42:48 UTC

I notice that a lot of you guys seem to be running a whole mess of distributed computing projects at once. Are any of you running more than 3 projects on one system? How is that effective? Isnt that spreading yourselves a little thin?

I run Milkyway@Home which seems to favor my GPU and also WCG which is working on the Childhood cancer project. And still sometimes they pause because something else may or may not be going on.

Although my system is currently a single core system you guys may all be running something far beefier.

AMD Athlon 3800+ CPU.
A8N-SLI Motherboard.
Nvidia 450GTS GPU.
4 GB DDR memory.

ID: 47148 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Bill Walker

Send message
Joined: 19 Aug 09
Posts: 23
Credit: 631,303
RAC: 0
Message 47156 - Posted: 7 Apr 2011, 19:59:41 UTC
Last modified: 7 Apr 2011, 20:02:50 UTC

Since my GPU machine went wonky, I'm only running a dual core laptop, with 4 projects (MW, S@H, WCG and Rosetta). Partly because I'm interested in lots of different projects, partly because I've seen two or even three projects run out of work at the same time. RAC per project is low, total RAC is relatively low (WCG and Rosetta aren't as good as MW in credits/hour), but who cares?
ID: 47156 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Chris Skull
Avatar

Send message
Joined: 16 Dec 10
Posts: 46
Credit: 205,697,511
RAC: 0
Message 47157 - Posted: 7 Apr 2011, 20:08:45 UTC

run so many diff project at once effectively?

I run 5 project at once on my quad-core... so in 5 hours time every project runs 4 hours and wait one hour... i don't think this is inefficient ?
The problem is, i love all this projects... and i MUST run them all :)


ID: 47157 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
BarryAZ

Send message
Joined: 1 Sep 08
Posts: 520
Credit: 302,525,188
RAC: 0
Message 47158 - Posted: 7 Apr 2011, 21:43:52 UTC - in response to Message 47148.  

On my various BOINC stations I run from 2 to 6 CPU projects and 1 to 3 (2 now with Dnetc out of the picture) GPU projects.

Nearly all the workstations are quads (one six core, a couple of dual core).

I find that with BOINC projects being what they are, having only one project can get problematic in terms of outages of one form or another.

That being said, I do find the need for some manual juggling (example -- Aqua runs multi-core, and tends to cause problems -- when it isn't queued to run but not suspended, it can take a quad core down to dual core for BOINC so I periodically suspend it to the let the other CPU projects play.

At the moment, since I am running a lot of HD 4850's - I have MW and Collatz to work with. Typically, that is something of an either/or as Collatz will tend to push MW out of the way to run so I will suspend Collatz to let MW process.


I notice that a lot of you guys seem to be running a whole mess of distributed computing projects at once. Are any of you running more than 3 projects on one system? How is that effective? Isnt that spreading yourselves a little thin?



ID: 47158 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Chris
Avatar

Send message
Joined: 3 Oct 10
Posts: 42
Credit: 320,242
RAC: 0
Message 47160 - Posted: 7 Apr 2011, 22:53:21 UTC

Thanks for the replies fellas. Very interesting. :)
32bit Windows XP Home
AMD Opteron 180
ASUS A8N-SLI Motherboard
Nvidia 450GTS GPU
4GB DDR Memory
ID: 47160 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Chris
Avatar

Send message
Joined: 3 Oct 10
Posts: 42
Credit: 320,242
RAC: 0
Message 47161 - Posted: 7 Apr 2011, 22:55:10 UTC - in response to Message 47158.  

I find that with BOINC projects being what they are, having only one project can get problematic in terms of outages of one form or another.


Good point. My system is currently only crunching away on the WCG project because MW@Home seems to be out of work for the time being.
32bit Windows XP Home
AMD Opteron 180
ASUS A8N-SLI Motherboard
Nvidia 450GTS GPU
4GB DDR Memory
ID: 47161 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Chris
Avatar

Send message
Joined: 3 Oct 10
Posts: 42
Credit: 320,242
RAC: 0
Message 47198 - Posted: 9 Apr 2011, 7:36:07 UTC - in response to Message 47161.  

Now that the board is back up, just wanted to let you guys know you have me running a bunch of diff projects now. I'm running MW@Home, WCG Help Fight Childhood Cancer and since the project crashed here I tried CFDN. Incredible program but shoved everything else out of the way. As much as climate prediction would be useful helping fight childhood cancer is way more urgent.

So I have suspended MW@Home and CFDN and have Help Fight Childhood Cancer going and Einstein@Home.
32bit Windows XP Home
AMD Opteron 180
ASUS A8N-SLI Motherboard
Nvidia 450GTS GPU
4GB DDR Memory
ID: 47198 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Bill Walker

Send message
Joined: 19 Aug 09
Posts: 23
Credit: 631,303
RAC: 0
Message 47531 - Posted: 11 Apr 2011, 1:07:56 UTC - in response to Message 47198.  
Last modified: 11 Apr 2011, 1:12:18 UTC

This has been discussed several times before here and over at S@H, but when you add a new project the BOINC scheduler appears to go bananas for a few days. I think it is trying to bring the new project up to its assigned percentage, and it will not run other projects that are closer to their assigned share than the new projects (which starts out with a history of zero % actual share). When you add a new project without adjusting the numeric value of the resource share of existing projects their % share goes down, so the scheduler thinks they are all over done at the moment (actual % will be over the new target %). Emergency mode keeps you from timing out on anything (at least in my experience), and it will all settle out in a few weeks.

Added in edit: you will also see this if a project goes down for a long time, and then comes back. Historical share has dropped, but target % remains, and the scheduler gives preference to whatever project is most out of whack.
ID: 47531 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote

Message boards : Number crunching : How are you guys able to run so many diff project at once effectively?

©2024 Astroinformatics Group