Page 1 of 1

Project: 2665 (Run 2, Clone 545, Gen 35)

Posted: Sun Sep 21, 2008 12:57 pm
by 314159
Linux Client, Stock Clock, Stable Q6600

Code: Select all

[11:14:55] - Calling './mpiexec -np 4 -host 127.0.0.1 ./FahCore_a1.exe -dir work/ -suffix 02 -checkpoint 15 -verbose -lifeline 23435 -version 602'

[11:14:55] 
[11:14:55] *------------------------------*
[11:14:55] Folding@Home Gromacs SMP Core
[11:14:55] Version 1.74 (November 27, 2006)
[11:14:55] 
[11:14:55] Preparing to commence simulation
[11:14:55] - Ensuring status. Please wait.
[11:14:56] - Starting from initial work packet
[11:14:57] 
[11:14:57] Project: 2665 (Run 2, Clone 545, Gen 35)
[11:14:57] 
[11:14:57] Assembly optimizations on if available.
[11:14:57] Entering M.D.
[11:15:13]  percent)
[11:15:14] - Starting from initial work packet
[11:15:14] 
[11:15:14] Project: 26Entering M.D.
[11:15:14] ne 545, Gen 35)
[11:15:14] 
[11:15:14] Entering M.D.
[11:15:22] Protein: HGG with glycExtra SSE boost OK.
[11:15:22] ocal files
[11:15:22] Extra SSE boost OK.
[11:30:24] t triggered.
[11:37:04] Writing local files
[11:37:04] Completed 2500 out of 250000 steps  (1 percent)

<snip>

[02:05:32] Completed 102500 out of 250000 steps  (41 percent)
[02:19:35] Warning:  long 1-4 interactions
[02:19:39] CoreStatus = 0 (0)
[02:19:39] Client-core communications error: ERROR 0x0
[02:19:39] Deleting current work unit & continuing...
[02:24:01] - Warning: Could not delete all work unit files (2): Core returned invalid code
[02:24:01] Trying to send all finished work units
[02:24:01] + No unsent completed units remaining.
Need some advice/assistance from one of our friendly Gurus (please):

I am an "individual" vs. "Corporate" folder.
Have 9 Quads running (Q6600's and faster), soon to be 11 (plus a "few" other machines). :ewink:
The Quads are being assigned WUs utilizing the A1 Core a bit less than half of the time.
I run only one SMP instance on each of these machines to preclude "extraneous" failures (perhaps being too conservative?).

I don't mind running occasional A1's on the Quads but these machines are obviously being substantially underutilized.

Could this be a function of my client configuration?

Can I "safely" run two SMP instances per Quad?
(I know that there is a thread addresssing this issue around here somewhere but cannot locate it. :(

Thank you in advance for any assistance! :)

Re: Project: 2665 (Run 2, Clone 545, Gen 35)

Posted: Sun Sep 21, 2008 2:51 pm
by toTOW
There's two other reports for Project: 2665 (Run 2, Clone 545, Gen 35) in the DB : one for 0 points, and one for partial credit.

This is probably a bad WU.

Re: Project: 2665 (Run 2, Clone 545, Gen 35)

Posted: Wed Sep 24, 2008 9:09 am
by bruce
314159 wrote:I don't mind running occasional A1's on the Quads but these machines are obviously being substantially underutilized.

Could this be a function of my client configuration?
I doubt there's anything that can be done about it. The A2 core is supposed to be the "cat's meow" when it comes to utilization, but it's still not considered stable enough to port to Windows. There are a lot of conflicting priorities that need attention by the Pande Group and this one is progressing slower than anybody would like. In the meantime, researchers are often forced to start new projects in a way that can make use of the larger population of Windows machines . . . .

Re: Project: 2665 (Run 2, Clone 545, Gen 35)

Posted: Wed Sep 24, 2008 12:13 pm
by 314159
Thanks Bruce but all of the Quads are now and have always been running Linux. :)

As I write this, there are 10 active (hopefully 11 by this afternoon).

Of the 10, 9 have been assigned WUs with the A1 Core. :( <---"sad" not "mad". :wink:

This is a substantial change over the past couple of days and I was curious concerning whether the A2 Core WUs have been temporarily pulled (or whether it is just a matter of Work Server weighting).

I do recall seeing a thread around here concerning running two SMP instances but cannot for the life of me locate it.
Would you kindly point this old man in the right direction?

Thanks, as always,
John

Edit: I totally missed your main point when I wrote the above. :oops:
I suspect that this is because there has always been a "proper" supply of A2's for the Linux platform and I just assumed that the projects that began with A2 would still be active.
All Quads are Win/Linux dual OS. If running two Win A1 instances is feasible I can look at a Win thread discussing the pros and cons.