Posts by gomeyer

1) Message boards : Number crunching : new workunit queue size (6) (Message 13212) Posted 28 Feb 2009 by gomeyer Post: Do you have the option to limit DL's to perhaps 10 at once but keep the total limit at 20 per core? That should ease the hits on the scheduler yet leave us with a little better comfort level if we do manage get a full quota, and let us get a partial load to hold us over in the mean time.
2) Message boards : Number crunching : new workunit queue size (6) (Message 13192) Posted 28 Feb 2009 by gomeyer Post: As of 1:44:14 UTC server status shows 0 results ready to send and the validator seems to have stopped. Rats, I may have to temporarily go back to SETI on these two machines. Noooooooo
3) Message boards : Number crunching : new workunit queue size (6) (Message 13074) Posted 27 Feb 2009 by gomeyer Post: Yup, the new limit came down right away on two machines with one manual update. Looks good so far.
4) Message boards : Number crunching : No work (Message 12954) Posted 26 Feb 2009 by gomeyer Post: Proof that Murphy is alive and well and living at MW: the only time my machine seems to run out of work is overnight when I'm asleep and can't do anything about it. @Travis, I don't know much about the BOINC system at your end, BUT . . (have you noticed that there is always a BUT in this type of statement?) I seem to remember that Matt Lebofsky (one of the admins over at Seti) once explained that work going out actually comes from a "feeder" which holds a small cache of wu's from the large store of those available for download. Even with plenty of work available, if this small feeder cache runs out the scheduler cannot send more work. He likened it to a single cashier in a store running out of change even though there was plenty of cash in the store's vault. With the new very fast GPU clients asking for big gobs of work, is it possible that the feeder is not "looking" at the "vault" often enough to stay full, or perhaps that small feeder cache needs to be increased a bit if possible? (I'm not knocking GPU processing, just pointing at a possible cause of the new phenomenon.) I imagine that this along with much else I don't understand is already being looked into, but just a thought . . .
5) Message boards : Number crunching : No work (Message 12561) Posted 23 Feb 2009 by gomeyer Post: Maybe it's (also?) a BOINC manager issue? . . . Good thought; BOINC is always one of the "usual suspects". In this case however I doubt it is the main cause as I am seeing the same thing with the old dependable 5.10.45.