Message boards :
Number crunching :
Shorties estimate up from three minutes to six hours after today's outage!
Message board moderation
Previous · 1 . . . 3 · 4 · 5 · 6 · 7 · 8 · 9 · Next
Author | Message |
---|---|
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13722 Credit: 208,696,464 RAC: 304 |
The longterm fix is really about a statistical problem - whether or not you include outlying datapoints into your calculations. They call them that for a reason. Which is why they are usually discarded when analysing data. Of course you first have to make sure that they are actually outliers & not actually significant data themselves. Statistics can be wonderfully manipulated by careful selection of outliers... Massage the data enough & you will always get the result you want. Grant Darwin NT |
Iona Send message Joined: 12 Jul 07 Posts: 790 Credit: 22,438,118 RAC: 0 |
The conditions you describe, regarding 'early exit', are exactly what happened with my ATI GPU WUs, last month....approximately half of the first ten WUs exited early (overflow results). Hence, these error values were, as I later suspected, being regarded as a very rapid 'full time completion', with the end result being that my GPU MB APR is in the order of almost 1400! Subsequent completion times for down-loaded WUs for the GPU (according to BOINC) were roughly as follows.... 2 mins for a 'shortie' and 6 mins for a 'normal' WU! Thats pretty impressive for any GPU, let alone a two-year-old ATI 4890; of course, any subsequent GPU WUs that I got, 'errored out' (-177). I then did a detatch and did a re-install of Lunatics 0.38, but without the MB app. It has only been in the last two or three days that I've run GPU WUs again, as I'd spotted the rather inflated timings on new CPU WUs - there was some chance to balance things, without resorting to 'rescheduling' and I've completed several WUs without error. For me, this may work out, but I won't be holding my breath for too long! Question. Probably a silly one, but, why is an approximate WU completion time needed? Don't take life too seriously, as you'll never come out of it alive! |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
The conditions you describe, regarding 'early exit', are exactly what happened with my ATI GPU WUs, last month....approximately half of the first ten WUs exited early (overflow results). Hence, these error values were, as I later suspected, being regarded as a very rapid 'full time completion', with the end result being that my GPU MB APR is in the order of almost 1400! Subsequent completion times for down-loaded WUs for the GPU (according to BOINC) were roughly as follows.... 2 mins for a 'shortie' and 6 mins for a 'normal' WU! Thats pretty impressive for any GPU, let alone a two-year-old ATI 4890; of course, any subsequent GPU WUs that I got, 'errored out' (-177). I then did a detatch and did a re-install of Lunatics 0.38, but without the MB app. It has only been in the last two or three days that I've run GPU WUs again, as I'd spotted the rather inflated timings on new CPU WUs - there was some chance to balance things, without resorting to 'rescheduling' and I've completed several WUs without error. For me, this may work out, but I won't be holding my breath for too long! To maintain a queue of work. If there was no estimated there would be no way determine how much work to ask for or how much to keep on to keep N days worth of work. My slowest machine has an estimated time of 72 hours. Which is about 3 days. So it keep 3-4 tasks on hand at a time for a 10 day queue. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
Iona Send message Joined: 12 Jul 07 Posts: 790 Credit: 22,438,118 RAC: 0 |
Yes, true. Perhaps I believe that a tad more info, should be kept on the 'client side' of things. Is this not a case of 'the left hand not knowing what the right hand is doing'? Don't take life too seriously, as you'll never come out of it alive! |
W-K 666 Send message Joined: 18 May 99 Posts: 19014 Credit: 40,757,560 RAC: 67 |
Question. Probably a silly one, but, why is an approximate WU completion time needed? If the benchmarks were accurate. If the estimated ops for each BOINC task was known. Then the estimated completion time could be calculated fairly accurately. But they are not so they use averages of every task that validates, which as we know is a flawed, because it includes premature exits. That causes the DCF to fluctuate wildly, because wrongly, there is only one DCF per project and not one per application. But that assumes the DCF needs to be there at all, which it probably isn't, because the estimates are now server side calculations. Scores out of 10 for; Design............ _______ Implementation....._______ Testing............_______ Fixing bugs........_______ (negative scores will not be rejected) |
Floyd Send message Joined: 19 May 11 Posts: 524 Credit: 1,870,625 RAC: 0 |
Question. Probably a silly one, but, why is an approximate WU completion time needed? Could they not get a pretty close estimate based on type of GPU and CPU and then err on the side of caution ? That would ( it seems ) give us a pretty regular amount of work in our cache... Maybe not as much as some want each time , but enough to at least keep the system stable. ??? |
kittyman Send message Joined: 9 Jul 00 Posts: 51468 Credit: 1,018,363,574 RAC: 1,004 |
I doubt it.... Too much variation even on identical resources depending on computer configuration, level of overclocking, the app being used, etc., etc., etc.. "Freedom is just Chaos, with better lighting." Alan Dean Foster |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
Question. Probably a silly one, but, why is an approximate WU completion time needed? Maybe that is what the Top GPU models list is going to help them with. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
Floyd Send message Joined: 19 May 11 Posts: 524 Credit: 1,870,625 RAC: 0 |
Ok... I understand... Even if we had a general Baseline WU time , and The ( I guess ) app info file was adjusted to let each one increase the work to a higher amount by say 10 % incraments at a time untill they got something that works for their own machine ??? I'm no way a programmer by any means... I have no Idea what it would take... These are just Ideas to run by those that have that kind of Knowledge in the field... |
jason_gee Send message Joined: 24 Nov 06 Posts: 7489 Credit: 91,093,184 RAC: 0 |
I doubt it.... LoL. Engineering worked out principles that led to 'Control Systems Theory' hundreds of years ago. Apparently Berkeley's library doesn't contain technical papers that old. Jason "Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions. |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
But that assumes the DCF needs to be there at all, which it probably isn't, because the estimates are now server side calculations. On this project, yes. But not every BOINC project uses the server-side 'CreditNew' code yet. Updating BOINC is a voluntary activity for project admins, too: any time there is a suggestion that projects have to upgrade their servers to support a particular feature 'because David says so', there is a howl of protest from users. DCF can only be removed from BOINC clients when the last active project server has upgraded to CreditNew. Predictions for when that will happen, anyone? But it gives me an idea for a feature suggestion for David: if a project is running server-side DCF, it sends a flag to (compatible) clients to say "lock your DCF to 1.0000: we're handling it at this end". Suggestions on a postcard to the usual place? |
kittyman Send message Joined: 9 Jul 00 Posts: 51468 Credit: 1,018,363,574 RAC: 1,004 |
Fat chance, that. The #6 rig on the project, my Frozen 920, has 2 GTX295s. And that GPU isn't even on the silly list. "Freedom is just Chaos, with better lighting." Alan Dean Foster |
Richard Haselgrove Send message Joined: 4 Jul 99 Posts: 14650 Credit: 200,643,578 RAC: 874 |
Apparently Berkeley's library doesn't contain technical papers that old. They don't call America the "New World" for nothing. |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
IIRC from the discussion about it earlier it only uses data from the stock app(s) & your machine is #5 now anyway. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
kittyman Send message Joined: 9 Jul 00 Posts: 51468 Credit: 1,018,363,574 RAC: 1,004 |
Ahhh.....ignoring the anonymous platforms again...LOL. And, thanks for the update. "Freedom is just Chaos, with better lighting." Alan Dean Foster |
W-K 666 Send message Joined: 18 May 99 Posts: 19014 Credit: 40,757,560 RAC: 67 |
But that assumes the DCF needs to be there at all, which it probably isn't, because the estimates are now server side calculations. It could be good idea, when these glitchs are resolved, to call the next release BOINC2, and with some good marketing. The general public always knows that the new all dancing version 2 has to be better than version 1, so therefore they will demand it is used. |
Dave Stegner Send message Joined: 20 Oct 04 Posts: 540 Credit: 65,583,328 RAC: 27 |
Can one of the guru's tell me if the below is another anomaly of the recent changes.? Notice the machine asks for 700K seconds of work, it only has 5 days of a 10 day cache, and the next request it asks for 0 seconds. Then 700K then 0. Also the responses are different, sometimes it simply says you got nothing and sometimes it responds about nothing for AP, etc. Aside from the fact that it is not receiving work, even though the Cricket graph is not maxed and work is available something else appears to be going on. Note nothing has changed on the machine for over 1 year. slws007.Staufferslandscape.com SETI@home 9/16/2011 10:41:54 AM Sending scheduler request: To fetch work. Requesting 754964 seconds of work, reporting 0 completed tasks slws007.Staufferslandscape.com SETI@home 9/16/2011 10:41:59 AM Scheduler request succeeded: got 0 new tasks slws007.Staufferslandscape.com SETI@home 9/16/2011 10:41:59 AM Message from server: No tasks sent slws007.Staufferslandscape.com SETI@home 9/16/2011 10:41:59 AM Message from server: No tasks are available for Astropulse v5 slws007.Staufferslandscape.com SETI@home 9/16/2011 10:41:59 AM Message from server: No tasks are available for Astropulse v505 slws007.Staufferslandscape.com SETI@home 9/16/2011 10:41:59 AM Message from server: No tasks are available for the applications you have selected. slws007.Staufferslandscape.com SETI@home 9/16/2011 11:00:35 AM Sending scheduler request: Requested by user. Requesting 0 seconds of work, reporting 0 completed tasks slws007.Staufferslandscape.com SETI@home 9/16/2011 11:00:40 AM Scheduler request succeeded: got 0 new tasks slws007.Staufferslandscape.com SETI@home 9/16/2011 11:02:01 AM Sending scheduler request: Requested by user. Requesting 756989 seconds of work, reporting 0 completed tasks slws007.Staufferslandscape.com SETI@home 9/16/2011 11:02:06 AM Scheduler request succeeded: got 0 new tasks slws007.Staufferslandscape.com SETI@home 9/16/2011 11:02:06 AM Message from server: Project has no tasks available slws007.Staufferslandscape.com SETI@home 9/16/2011 11:03:15 AM Sending scheduler request: Requested by user. Requesting 0 seconds of work, reporting 0 completed tasks slws007.Staufferslandscape.com SETI@home 9/16/2011 11:03:20 AM Scheduler request succeeded: got 0 new tasks slws007.Staufferslandscape.com SETI@home 9/16/2011 11:06:35 AM Sending scheduler request: Requested by user. Requesting 0 seconds of work, reporting 0 completed tasks slws007.Staufferslandscape.com SETI@home 9/16/2011 11:06:40 AM Scheduler request succeeded: got 0 new tasks slws007.Staufferslandscape.com SETI@home 9/16/2011 11:11:45 AM Sending scheduler request: To fetch work. Requesting 758268 seconds of work, reporting 0 completed tasks slws007.Staufferslandscape.com SETI@home 9/16/2011 11:11:50 AM Scheduler request succeeded: got 0 new tasks slws007.Staufferslandscape.com SETI@home 9/16/2011 11:11:50 AM Message from server: No tasks sent slws007.Staufferslandscape.com SETI@home 9/16/2011 11:11:50 AM Message from server: No tasks are available for Astropulse v5 slws007.Staufferslandscape.com SETI@home 9/16/2011 11:11:50 AM Message from server: No tasks are available for Astropulse v505 slws007.Staufferslandscape.com SETI@home 9/16/2011 11:11:50 AM Message from server: No tasks are available for the applications you have selected. Dave |
Dave Stegner Send message Joined: 20 Oct 04 Posts: 540 Credit: 65,583,328 RAC: 27 |
Not to beat a dead horse but, the previous posting was 11 am and it is currently (still) acting that way. Host Project Date Message slws007.Staufferslandscape.com SETI@home 9/16/2011 3:23:57 PM Sending scheduler request: To fetch work. Requesting 788142 seconds of work, reporting 0 completed tasks slws007.Staufferslandscape.com SETI@home 9/16/2011 3:24:02 PM Scheduler request succeeded: got 0 new tasks slws007.Staufferslandscape.com SETI@home 9/16/2011 3:24:02 PM Message from server: No tasks sent slws007.Staufferslandscape.com SETI@home 9/16/2011 3:24:02 PM Message from server: No tasks are available for Astropulse v5 slws007.Staufferslandscape.com SETI@home 9/16/2011 3:24:02 PM Message from server: No tasks are available for Astropulse v505 slws007.Staufferslandscape.com SETI@home 9/16/2011 3:24:02 PM Message from server: No tasks are available for the applications you have selected. slws007.Staufferslandscape.com SETI@home 9/16/2011 3:29:07 PM Sending scheduler request: To fetch work. Requesting 788747 seconds of work, reporting 0 completed tasks slws007.Staufferslandscape.com SETI@home 9/16/2011 3:29:12 PM Scheduler request succeeded: got 0 new tasks slws007.Staufferslandscape.com SETI@home 9/16/2011 3:29:12 PM Message from server: No tasks sent slws007.Staufferslandscape.com SETI@home 9/16/2011 3:29:12 PM Message from server: No tasks are available for Astropulse v5 slws007.Staufferslandscape.com SETI@home 9/16/2011 3:29:12 PM Message from server: No tasks are available for Astropulse v505 slws007.Staufferslandscape.com SETI@home 9/16/2011 3:29:12 PM Message from server: No tasks are available for the applications you have selected. slws007.Staufferslandscape.com SETI@home 9/16/2011 3:30:55 PM Sending scheduler request: Requested by user. Requesting 0 seconds of work, reporting 0 completed tasks slws007.Staufferslandscape.com SETI@home 9/16/2011 3:31:00 PM Scheduler request succeeded: got 0 new tasks slws007.Staufferslandscape.com SETI@home 9/16/2011 3:36:08 PM Sending scheduler request: To fetch work. Requesting 788242 seconds of work, reporting 0 completed tasks slws007.Staufferslandscape.com SETI@home 9/16/2011 3:36:13 PM Scheduler request succeeded: got 0 new tasks slws007.Staufferslandscape.com SETI@home 9/16/2011 3:36:13 PM Message from server: No tasks sent slws007.Staufferslandscape.com SETI@home 9/16/2011 3:36:13 PM Message from server: No tasks are available for Astropulse v5 slws007.Staufferslandscape.com SETI@home 9/16/2011 3:36:13 PM Message from server: No tasks are available for Astropulse v505 slws007.Staufferslandscape.com SETI@home 9/16/2011 3:36:13 PM Message from server: No tasks are available for the applications you have selected. slws007.Staufferslandscape.com SETI@home 9/16/2011 3:39:07 PM Sending scheduler request: Requested by user. Requesting 0 seconds of work, reporting 0 completed tasks slws007.Staufferslandscape.com SETI@home 9/16/2011 3:39:12 PM Scheduler request succeeded: got 0 new tasks Dave |
Wembley Send message Joined: 16 Sep 09 Posts: 429 Credit: 1,844,293 RAC: 0 |
But it gives me an idea for a feature suggestion for David: if a project is running server-side DCF, it sends a flag to (compatible) clients to say "lock your DCF to 1.0000: we're handling it at this end". Suggestions on a postcard to the usual place? Write your suggestions on a piece of toilet paper and flush it with the rest of them, has just as much a chance of David listening as any other way. |
kittyman Send message Joined: 9 Jul 00 Posts: 51468 Credit: 1,018,363,574 RAC: 1,004 |
But it gives me an idea for a feature suggestion for David: if a project is running server-side DCF, it sends a flag to (compatible) clients to say "lock your DCF to 1.0000: we're handling it at this end". Suggestions on a postcard to the usual place? Now, that's not completely true. Granted, Dr. Anderson does have his way of going ahead with his agenda without notifying anybody...as I guess is his right. But he has also taken bug reports and suggestions and put in fixes for them on occasion as well. Please don't take this as meaning that I agree with everything he does....the current thread topic would be a definite disagreement. I just think the comment made was a bit harsh. Meow. "Freedom is just Chaos, with better lighting." Alan Dean Foster |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.