Message boards :
Number crunching :
Can't maintain tasks cache
Message board moderation
Author | Message |
---|---|
Send message Joined: 27 Jun 09 Posts: 85 Credit: 39,805,338 RAC: 0 |
When I joined to MW my host can hold only 24 MW tasks (it's quad ). When BOINC requested more it recived answer from server like "quota of 6 results per CPU reached" 24 tasks is pretty small cache when each task computes just one minute. But it was OK while server online. But Now MW sends tasks on BOINC requests only when there are no MW tasks in cache at all. It leads to some delay (all tasks already computed, but new ones still need to be requested and downloaded). Cause currently MW is the single project known to me with ATI GPUs support, there are nothing available to fill pauses in GPU work. Does MW server overloaded now and can't keep up with work requests or some tuning in project shares or in another settings should be done? Example of BOINC log: 09/07/2009 01:41:53 Milkyway@home Sending scheduler request: To fetch work. 09/07/2009 01:41:53 Milkyway@home Requesting new tasks 09/07/2009 01:41:58 Milkyway@home Scheduler request completed: got 0 new tasks 09/07/2009 01:42:59 Milkyway@home Sending scheduler request: To fetch work. 09/07/2009 01:42:59 Milkyway@home Requesting new tasks 09/07/2009 01:43:05 Milkyway@home Scheduler request completed: got 0 new tasks 09/07/2009 01:43:39 Milkyway@home Computation for task ps_sgr_210F5_2s_hiw_14393531_1247089230_0 finished 09/07/2009 01:43:39 Milkyway@home Computation for task ps_sgr_210F5_2s_hiw_14393532_1247089230_0 finished 09/07/2009 01:43:39 Milkyway@home Computation for task ps_sgr_210F5_2s_hiw_14393533_1247089230_0 finished 09/07/2009 01:44:07 Milkyway@home Sending scheduler request: To fetch work. 09/07/2009 01:44:07 Milkyway@home Reporting 3 completed tasks, requesting new tasks 09/07/2009 01:44:12 Milkyway@home Scheduler request completed: got 0 new tasks 09/07/2009 01:45:13 Milkyway@home Sending scheduler request: To fetch work. 09/07/2009 01:45:13 Milkyway@home Requesting new tasks 09/07/2009 01:45:18 Milkyway@home Scheduler request completed: got 0 new tasks 09/07/2009 01:46:19 Milkyway@home Computation for task ps_sgr_212F5_2s_hiw_14392128_1247089048_1 finished 09/07/2009 01:46:19 Milkyway@home Computation for task ps_sgr_210F5_2s_hiw_14393515_1247089229_0 finished 09/07/2009 01:46:19 Milkyway@home Sending scheduler request: To fetch work. 09/07/2009 01:46:19 Milkyway@home Requesting new tasks 09/07/2009 01:46:25 Milkyway@home Scheduler request completed: got 0 new tasks 09/07/2009 01:46:42 Milkyway@home Computation for task ps_sgr_210F5_3s_hiw_14393553_1247089230_0 finished 09/07/2009 01:47:05 Milkyway@home Computation for task ps_vrfy15_3s_1_14391998_1247089004_1 finished 09/07/2009 01:47:28 Milkyway@home Sending scheduler request: To fetch work. 09/07/2009 01:47:28 Milkyway@home Reporting 4 completed tasks, requesting new tasks 09/07/2009 01:47:33 Milkyway@home Scheduler request completed: got 8 new tasks (that is, server gave only 8 tasks and only after all prev results were reported). [ tried solution from this thread http://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=971 If caoonect every.. setted to 3 days MW tasks go into high priority mode. I leaved this setting with 2 days value. But still no additional MW tasks in queue, only that wich running already... ] |
Send message Joined: 12 Nov 07 Posts: 2425 Credit: 524,164 RAC: 0 |
You're running other projects on that computer right? I would say it's trying to allow each project to do its share. As Mw built up it's allowing the others their share now. Doesn't expecting the unexpected make the unexpected the expected? If it makes sense, DON'T do it. |
Send message Joined: 27 Jun 09 Posts: 85 Credit: 39,805,338 RAC: 0 |
BOINC scheduler doesn't think so. If you look on provided log you will see that it requests new work from server. But server gives zero reply. So, server knows better what project shares my PC has than its own BOINC client? |
©2024 Astroinformatics Group