| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: FightAIDS@Home
Thread: Problems with progress calculation |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 29
|
|
| Author |
|
|
TOMinAZ
Cruncher United States Joined: Feb 11, 2007 Post Count: 40 Status: Offline Project Badges: 
|
Well, the last work unit that finished ran for nearly 6 days, with maybe only about 12 hours interruption. By my calculation, it took about 110 hours to actually run, and I got credit for only 61.+ hours.
I've actually timed it while running. After exactly 4 hours of running, it only showed an increase of just slightly less than 2 hours of run time. When I see 'time estimated to completion', I usually double that time and I know that 40 or 50 hours of run time will actually take almost a week to complete. I guess BOINC's clock runs slower than realtime. I thought maybe it was just my machine. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
TOMinAZ, the time you get credit for is CPU time, not wall clock time. The discrepancy you report almost exactly matches the throttle setting of 60%. If you want to contribute more, and finish work faster, then you can select the "Maximum Output" option in your device profile.
|
||
|
|
Rickjb
Veteran Cruncher Australia Joined: Sep 17, 2006 Post Count: 666 Status: Offline Project Badges: 
|
RobertM, and others confused/worried about the long FAAH WUs and erratic estimates of Completion Time:
----------------------------------------Don't worry, everything is under control (mostly). See knreed's post in Known Issues: Long Running FightAIDS@Home Workunits These long WUs are for the new Experiment 24, which has been given high priority. See the post by mgl_ALPerryman (FightAIDS@Home Scientist) in thread "experiment confusion" and the updated FA@H Status Page It seems that we are helping to investigate how a real HIV mutant resists real protease inhibitor drugs. When people ask you what practical good has come from your grid computing efforts, you can show this to them! In another thread, it was explained that the % Progress of a FAAH or DDDT WU sometimes decreases by a small amount when the application backtracks and tries a new position for a previously-positioned atom. Also, BOINC's Completion Time estimates have recently been thoroughly confused by the huge fluctuations in WU length. The combination of these effects probably explains what you've been seeing. Don't worry, be happy  Update: A problem has been found in the way these very long WUs were generated. See lawrencehardin's post in the thread VERY LONG Work Unit!  [Edit 2 times, last edit by Rickjb at Aug 2, 2008 10:40:39 AM] |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Rickjb, Experiment 24 to be correct.
----------------------------------------
WCG
Please help to make the Forums an enjoyable experience for All! |
|||
|
||||
|
TOMinAZ
Cruncher United States Joined: Feb 11, 2007 Post Count: 40 Status: Offline Project Badges:
|
TOMinAZ, the time you get credit for is CPU time, not wall clock time. The discrepancy you report almost exactly matches the throttle setting of 60%. If you want to contribute more, and finish work faster, then you can select the "Maximum Output" option in your device profile. Of course! How silly of me. I had forgotten about the CPU throttle. I guess the difference becomes more noticeable with longer runtimes. Thanks! |
|||
|
||||
|
robertmiles
Senior Cruncher US Joined: Apr 16, 2008 Post Count: 445 Status: Offline Project Badges: 
|
robertmiles, Assuming that to you the FA@H project is synonymous to WCG, you really don't have to do that suspending and it's better not to do so. BOINC keeps a Long and Short Term Debt record on seconds crunched for each project. If WCG due the long job gets too much crunching time in now, it will stop fetching work from WCG and will let the other project have a go until the balance sheet is fairly even again. ciao For now, I'll let WCG continue, but not for the FA@H project. After the burst of workunits expected from the non-WCG project, I'll let FA@H resume. |
||
|
|
robertmiles
Senior Cruncher US Joined: Apr 16, 2008 Post Count: 445 Status: Offline Project Badges:
|
Now, another faah workunit has a similar problem, but not as bad. 7/30/2008 10:31:10 AM|World Community Grid|Resuming task faah5012_1hef_1hpx_00_0 using faah version 605 If I remember correctly, this one started out with an estimate of about 44 hours required. It's now approaching 12 hours CPU time, with a progress of about 15%, with an estimate of about 41 hours to completion. It's now approaching 22 hours CPU time and 29% Progress, with a To completion time of over 37 hours. The CPU time seems to advance a few seconds for every second of drop in the To completion time, so I suspect that the 44 hours estimated time was still a significant underestimate. It's now at 39 hours CPU time, 51% Progress, and 29 hours To completion. Since I'm expecting a burst of workunits from another project, I've temporarily disabled getting more workunits from the faah project. It's now at 61 hours CPU time, 81% progress, and 13 hours to completion. At least the time to completion is decreasing about as much as the CPU time is increasing. |
||
|
|
robertmiles
Senior Cruncher US Joined: Apr 16, 2008 Post Count: 445 Status: Offline Project Badges:
|
Now, it errored out due to exceeding the CPU time limit (235228) and only reported the use of 22.39 hours CPU time. Someone else is running another copy now, and may have the same result in a few days.
|
||
|
|
robertmiles
Senior Cruncher US Joined: Apr 16, 2008 Post Count: 445 Status: Offline Project Badges:
|
Is the status "Ready to Report" ? If so, this is normal. When you finish crunching a WU, BOINC contacts the server and uploads the data but it only does this at certian times. Results will be reported automatically at the following triggers: 1. New work is needed. Completed work will be reported at the same time. 2. The Deadline is less than 24 hours away. 3. The Result is less than the Work Buffer setting from it's Deadline 4. The Result has been in the Ready to Report status for 24 hours. You can force the upload if required by going into Boinc manager and selecting the WU and then advanced->retry communications. If you then look in Messages, you should see something like: 29/07/2008 08:29:11|World Community Grid|Sending scheduler request: To report completed tasks. Requesting 0 seconds of work, reporting 1 completed tasks 29/07/2008 08:29:16|World Community Grid|Scheduler request succeeded: got 0 new tasks If you still have problems, please let us know what you have done and any messages that are shown. I just had another such workunit caught at 100%. The status was ready to run, so it apparantly wasn't quite complete yet. I suspended a workunit with the status running, and the workunit with the problem took over and completed in a short time. 8/12/2008 5:39:01 PM|World Community Grid|Resuming task faah4295_indazoleOH_DiMthxbnzMIN_xmd09230_02_1 using faah version 605 8/12/2008 5:39:04 PM|World Community Grid|Computation for task faah4295_indazoleOH_DiMthxbnzMIN_xmd09230_02_1 finished |
||
|
|
|