Questions and Answers :
Windows :
Time to completion?
Message board moderation
Author | Message |
---|---|
shawn Send message Joined: 19 Jun 02 Posts: 9 Credit: 407,589 RAC: 0 |
My machine takes about 3-1/2 hours to complete a work unit. I noticed that after the site came back up on Jul 9th that the client will only download one work unit at a time. The work unit now shows an estimated time till completion of 22-1/2 hours. As soon as it starts crunching it does get corrected. Any thoughts? |
SkyHook Send message Joined: 20 Feb 02 Posts: 9 Credit: 194,866 RAC: 0 |
Depending on who's story you care to believe, the Time To Complete for WUs has either accidentally or purposely been inflated. The end result either way is that the load on the distribution system has been relaxed so that most if not all people are getting some WUs, just not as many as they may have liked to. You will find that although the projected time is inflated, these WUs still actually crunch within the time span you were accustomed to seeing. SkyHook |
MPBroida Send message Joined: 6 Sep 00 Posts: 337 Credit: 16,433 RAC: 0 |
Whatever provides those "pre-download estimates" is WAY off for the newer units. The values are ludicrous. And it doesn't really get "corrected" when it starts crunching; the run-time estimator is (and has been since they went live), even worse than the pre-download estimator is now. EVERY unit I get runs to over 50% in less than 10 minutes, but the estimated time-to-completion keeps going UP. It's nowhere near accurate until it shows at least 85% complete. Units on this system take 7-8 hours; on my other one they take 3-4 hours. So saying 50% after 10 minutes is just silly. They KNOW approximately how many FFTs a "normal" unit is going to have to run, and they have a benchmark of our system speed. That should provide a "reasonable" estimate right from the start of the unit, though it will likely need some adjustment as it goes along. Their "time-to-completion" code in the client is messed up badly, making that info totally worthless. |
Rich[FL] Send message Joined: 10 Feb 00 Posts: 3 Credit: 2,608,254 RAC: 0 |
The problem this has caused as been the lack of downloads altogether. Because of testing on the BOINC Beta Test site, awhile back they recommended to us to reduce our cache limit to a very short time (0.1 - 0.5 days) so we could catch the software updates they were sending out. You probably remember that. I just caught a glimpse of another answer that led me to believe that may be my problem in getting work units for SETI. I signed up a long time ago here and have NEVER received work units. I decided to up my cache to 0.1 - 1.0 days (just set it at 2 days today) and I'm finally getting work units!!! So, that 18+hr crunch time estimate has been preventing me from even getting work units at all. I'm sure others have figured this out long ago (I'm kinda slow on the uptake because of my work schedule). It's just frustrating to not being able to participate in this endeavor for so long because of this little "glitch". Hope others who aren't getting WUs will see this and make the appropriate adjustments so they may start getting work. Rich |
Heffed Send message Joined: 19 Mar 02 Posts: 1856 Credit: 40,736 RAC: 0 |
> I just caught a glimpse of another answer that led me to believe that may be > my problem in getting work units for SETI. I signed up a long time ago here > and have NEVER received work units. I decided to up my cache to 0.1 - 1.0 > days (just set it at 2 days today) and I'm finally getting work units!!! So, > that 18+hr crunch time estimate has been preventing me from even getting work > units at all. Hey Rich, the inflated time to completion has only been an issue for a couple days. <a> [/url] |
shawn Send message Joined: 19 Jun 02 Posts: 9 Credit: 407,589 RAC: 0 |
> Depending on who's story you care to believe, the Time To Complete for WUs has > either accidentally or purposely been inflated. The end result either way is > that the load on the distribution system has been relaxed so that most if not > all people are getting some WUs, just not as many as they may have liked to. > > SkyHook > > > I did suspect the Admin may have set the completion time intentionally to ease the load on the servers. I made that assumption when I noticed they always arrive with the exact same estimated time to completion of 22:30:25. > You will find that although the projected time is inflated, these WUs still > actually crunch within the time span you were accustomed to seeing. No, it doesn't affect crunch time at all. Since I'm currently stuck with dial-up it means I have to do a lot of babysitting to keep the client busy. |
Mr. Kevvy Send message Joined: 15 May 99 Posts: 3797 Credit: 1,114,826,392 RAC: 3,319 |
I hope that this wasn't done on purpose... I can't get any Predictor@Home work because I have two SAH WU's sitting at the bottom that have an estimated completion time of 35 hours, which is 8-10 times longer than they take, so BOINC thinks I have enough WU's to last 3 days when it's more like 8 hours. The SAH WU's that I had prior have estimated completion times of 5:52. Now they are up to 35:12 which is exactly 6 times as long (to the second.... no rounding at all.) The SAH project code did not change during this time. This is affecting two machines (one WinXP on BOINC 3.20 and one Win2K on BOINC 3.19) so it doesn't seem like a BOINC problem or a SAH project code problem. |
MPBroida Send message Joined: 6 Sep 00 Posts: 337 Credit: 16,433 RAC: 0 |
> rounding at all.) The SAH project code did not change during this time. This > is affecting two machines (one WinXP on BOINC 3.20 and one Win2K on BOINC > 3.19) so it doesn't seem like a BOINC problem or a SAH project code problem. When did "BOINC 3.20" come out?? After the 6/22 announcement that they had "gone live", I downloaded and it was only 3.19. They never announced a 3.20 version. |
Thierry Van Driessche Send message Joined: 20 Aug 02 Posts: 3083 Credit: 150,096 RAC: 0 |
> When did "BOINC 3.20" come out?? After the 6/22 announcement that they had > "gone live", I downloaded and it was only 3.19. They never announced a 3.20 > version. You will find the answer in this thread. Greetings from Belgium. |
MPBroida Send message Joined: 6 Sep 00 Posts: 337 Credit: 16,433 RAC: 0 |
> > When did "BOINC 3.20" come out?? After the 6/22 announcement that they > had > > "gone live", I downloaded and it was only 3.19. They never announced a > 3.20 > > version. > > You will find the answer in <a> href="http://setiweb.ssl.berkeley.edu/forum_thread.php?id=1354">this > thread[/url]. Great. Thanks! :) |
Mouser Send message Joined: 22 May 02 Posts: 4 Credit: 85,199 RAC: 0 |
I had the same problem before, but after a few WU's were finished the Time to Completion became correct. At 1:58:00 |
©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.