Message boards :
Questions and problems :
BOINC 6.6.20 released for Windows, Windows x64, Linux, Linux x64 and MacOS X
Message board moderation
Previous · 1 · 2 · 3 · Next
Author | Message |
---|---|
Send message Joined: 5 Oct 06 Posts: 5081 ![]() |
I do not care how long the work units I get are, but I doubt that any SETI CUDA WUs should have an estimated competion time of 365 hours on a Q6600. Are the estimates any less 'wrong' now than when you first posted, over 12 hours ago? In other words, is BOINC starting to self-correct, as it is designed to? - though you would need to complete several dozen tasks on each project to get anywhere near complete correction. Also, are you allowing for the fact that v6.6.20 is now displaying (and estimating) "elapsed" time instead of "CPU" time previously. This makes a big difference for CUDA tasks, and it may make a difference whether you let BOINC compute all the time, or have it set to stop computing when the computer is in use for other tasks. |
Send message Joined: 21 Mar 09 Posts: 20 ![]() |
I do not care how long the work units I get are, but I doubt that any SETI CUDA WUs should have an estimated competion time of 365 hours on a Q6600. I have everything set to run at 100%. I have nothing else running on this machine it only runs BOINC and is not used for anything else. THe estimates for GPU and CPU are wrong. I understand this is elapsed time, but all of the work units are sowing an estimate of over 300 hours to completion, and are only taking 4 to 6 hours. It looks like while they are running the completion time reduces but never actaully really corrects itself. I have on WCG CPU task that shows 3:24:52 elapsed, 66% done, to completion 14:30:12 which makes no sense. When this started it actually shoed 482:57:33. As I have said, everything seems to be processing ok, in the expected time frames. My concern is that with the estimated completion times over 300 hours, BOINC has not downloaded any new SETI or WCG tasks since I did the upgrade. I have been running for 24 hours under 6.6.20 and there has been no change. I have about 6 SETI WUs and 17 WCG WUs in the queue, all with estimated completion times of over 300 hours, but with actual completion times projected to be 5 hours. I have about a days worth of work left for both. I supose that I can wait until I clean out my queue and then reset SETI and WCG. I am running 18 machines for BOINC and this is the only one that shows this behavior of totally inaccurate completion times. It happend after upgrading from 6.6.0 to 6.6.20. I have upgraded two other machines with no problems, but do not want to upgrade anymore until I figure this out. How does BOINC caculate the estimate to completion. Is there something I have set wrong that is confusing it? Is this a bug in 6.6.20? |
![]() Send message Joined: 20 Dec 07 Posts: 1069 ![]() |
How does BOINC calculate the estimate to completion. Is there something I have set wrong that is confusing it? Is this a bug in 6.6.20? I have no answer to your questions, but did you check the result of those calculations on the Details page of Your computers? Namely "Average CPU efficiency" and "Task duration correction factor". |
Send message Joined: 21 Mar 09 Posts: 20 ![]() |
[quote]How does BOINC calculate the estimate to completion. Is there something I have set wrong that is confusing it? Is this a bug in 6.6.20? I have no answer to your questions, but did you check the result of those calculations on the Details page of Your computers? Namely "Average CPU efficiency" and "Task duration correction factor".[/quote My duration correction factor is HUGE. Is this the problem! Task duration correction factor 86.759626. On all of my other computers it is less than 2. My task duration time for WCG on this computer is 74.820229. Can this cause incorrect comletion times? If it does, how do I reset it. I also do not understand "While BOINC running, % of time work is allowed 13.8538 %". This is 99.99% on my other computers. Here is my Detail for the machine, from SETI in question. CPU type GenuineIntel Intel(R) Core(TM)2 Quad CPU Q6600 @ 2.40GHz [x86 Family 6 Model 15 Stepping 11] Number of processors 4 Coprocessors NVIDIA GeForce 8500 GT (512MB) driver: 17824 Operating System Microsoft Windows XP Home x86 Editon, Service Pack 2, (05.01.2600.00) Memory 2046.82 MB Cache 244.14 KB Swap space 4336.64 MB Total disk space 14.65 GB Free Disk Space 13.18 GB Measured floating point speed 2701.35 million ops/sec Measured integer speed 5895.7 million ops/sec Average upload rate 54.59 KB/sec Average download rate 73.54 KB/sec Average turnaround time 3.47 days Maximum daily WU quota per CPU 100/day Tasks 27 Number of times client has contacted server 615 Last time contacted server 26 Apr 2009 14:24:01 UTC % of time BOINC client is running 99.8609 % While BOINC running, % of time host has an Internet connection 100 % While BOINC running, % of time work is allowed 13.8538 % Task duration correction factor 86.759626 |
![]() Send message Joined: 8 Jan 06 Posts: 448 ![]() |
You posted as I was preparing the information below. Task duration correction factor 86.759626 This way out of normal range. ------------------------ To estimate initial completion time, Boinc uses a value supplied by the project which is then multiplied by the DCF (duration correction factor) for that project to adjust for your system capability. A separate DCF value is maintained for each project. The DCF is adjusted at the completion of a WU. It goes up quickly if the estimate was too low, but is adjusted down more slowly if the estimate was too high. This is to ensure that you rarely download more work than the system can handle. The DCF will self correct but if it is way off this could take the completion dozens of WU from a project to adjust itself down. It does sounds like the your DCF may have been messed up. To check it out, in version 6.6.x, select the projects individually in the Projects tab and click properties button. The DCF entry is at the bottom of the properties display. If the value is above 2. then that is most likely the problem. Depending on the accuracy of project estimates and your system capability, normal range is between .3 and 2. The greater the value the larger initial estimate will be. For a quick fix you can manually edit the DCF entries in a plain text editor like notepad. The file that needs editing contains most of the parameters Boinc needs to handle the projects properly so you need to be confident you can do this without messing up. Before editing the file, make sure to stop Boinc and all running applications. Care most be taken to just change the value and nothing else. Also ensure that you 'Save' the file (not save as) so that it's not saved as a TXT file. The DCF entries are in the client_state.xml file in the Boinc Data folder. Each project has its own entry that will need to be located and changed. The entry should look something like <duration_correction_factor>1.833852</duration_correction_factor> or <duration_correction_factor>0.533852</duration_correction_factor> correct each projects that is way off with <duration_correction_factor>1.000000</duration_correction_factor> will reset to the default value. Note the 6 position decimal number. Boinc V 7.4.36 Win7 i5 3.33G 4GB NVidia 470 |
![]() Send message Joined: 29 Aug 05 Posts: 15480 ![]() |
If you don't feel like editing your client_state.xml file, just put the projects affected to No New Tasks, run their cache dry, upload & report all work and then reset the project. Resetting will delete all work in progress, as well as all applications, hence the advice to run your cache dry. Resetting will also set DCF to 1.0 |
Send message Joined: 21 Mar 09 Posts: 20 ![]() |
If you don't feel like editing your client_state.xml file, just put the projects affected to No New Tasks, run their cache dry, upload & report all work and then reset the project. Resetting will delete all work in progress, as well as all applications, hence the advice to run your cache dry. Resetting will also set DCF to 1.0 Thanks, you are all being helpful. I made the changes to <duration_correction_factor>1.000000</duration_correction_factor> for both SETI and WCG and it fixed the "to completion" times. They are now between 4 and 5 hours which seem more resonable. The factor probably should be a little more than 1 for this computer but I expect it will adjust. Unfortunately this did not seem to fix the problem of not being able to download. I noticed another strange number that does not make sense. While BOINC running, % of time work is allowed 16.879 % This computer runs 24x7 and does nothing but BOINC. On all my other computers this factor is 99% While BOINC running, % of time work is allowed 99.9915 % I still cannot get downloads and I get these messages from both projects. 4/26/2009 6:18:28 PM SETI@home Sending scheduler request: Requested by user. 4/26/2009 6:18:28 PM SETI@home Not reporting or requesting tasks 4/26/2009 6:18:33 PM SETI@home Scheduler request completed: got 0 new tasks 4/26/2009 6:18:53 PM World Community Grid Sending scheduler request: Requested by user. 4/26/2009 6:18:53 PM World Community Grid Not reporting or requesting tasks 4/26/2009 6:18:58 PM World Community Grid Scheduler request completed: got 0 new tasks Any other suggestions on how I might get downloads to work. I am going to run out of WUs in 24 hours. |
![]() Send message Joined: 8 Jan 06 Posts: 448 ![]() |
You could edit that also. Its near the top of the client_state.xml file. mine reads <time_stats> <on_frac>0.976760</on_frac> <connected_frac>0.985244</connected_frac> <active_frac>0.999907</active_frac> <last_update>1240788829.156250</last_update> </time_stats> Note: don't touch the time stamp. |
Send message Joined: 21 Mar 09 Posts: 20 ![]() |
You could edit that also. Its near the top of the client_state.xml file. Thanks, I am making progress. Changing the duration factor fixed the completion times for both SETI and WCG. Changing Active_frac as you suggested fixed my doneload problem for SETI. Once I made the change you suggested and I restarted BOINC, I down loaded a bunch of WUs for SETI. I am still not downloading anything for WCG. I still get this message: 4/26/2009 7:34:51 PM World Community Grid Sending scheduler request: Requested by user. 4/26/2009 7:34:51 PM World Community Grid Not reporting or requesting tasks 4/26/2009 7:34:56 PM World Community Grid Scheduler request completed: got 0 new tasks I did notice in Client_state.xml for WCG: <short_term_debt>-86400.000000</short_term_debt> <long_term_debt>-7508364.109212</long_term_debt> These seem to be large negative numbers. For Seti they are positive and zero <short_term_debt>86400.000000</short_term_debt> <long_term_debt>0.000000</long_term_debt> Not sure what this is. Any other ideas why I may still be having an issue downloading under WCG. Thanks, Mitch |
![]() Send message Joined: 8 Jan 06 Posts: 448 ![]() |
I don't understand the new system for long term debt other then the top value is now zero. There are only a couple of people that have a grasp of the new system. You could try setting WCG long term debt to zero, I don't think it will hurt anything too badly, but I'm not sure that it will have the immediate desired effect. I think everything else is probably in the correct range. A little patience may just be needed until boinc finds a new balance. |
Send message Joined: 21 Mar 09 Posts: 20 ![]() |
I think I will wait another day before I run out of WUs. If WUs go to zero it may start downloading again. If not I will try a reset. SETI is working fine right now. WCG has not download a work unit since the upgrade and I have just 16 left in the queue. With four CPUS it should run out tomorrow evening. - Mitch |
Send message Joined: 21 Mar 09 Posts: 20 ![]() |
I don't understand the new system for long term debt other then the top value is now zero. There are only a couple of people that have a grasp of the new system. You could try setting WCG long term debt to zero, I don't think it will hurt anything too badly, but I'm not sure that it will have the immediate desired effect. So I got impatient and could not wait. I reset long term debt to zero and it freed everything up. I downloaded a whole bunch of WUs. Everything seems to be working now. Still do not know why I needed to manually reset all of these things becasue of an upgrade. Something else must have happened. Thanks again for everyones help. I learned a lot :) _ Mitch |
![]() Send message Joined: 29 Aug 05 Posts: 15480 ![]() |
Still do not know why I needed to manually reset all of these things becasue of an upgrade. Something else must have happened. BOINC does not reset debts, even though the method of how these debts are used changed between 6.4 and all versions before that and 6.6 So it'll go with the old values, converts them to the new method and then gets very much stuck. I have requested twice now that they please release a client that resets debts when upgrading from a pre-BOINC 6.6 version, but was denied. Apparently everyone should be able to use the <zero_debts> flag once in cc_config.xml (sigh). |
Send message Joined: 21 Mar 09 Posts: 20 ![]() |
Still do not know why I needed to manually reset all of these things becasue of an upgrade. Something else must have happened. Except I upgraded from 6.6.0 to 6.6.20. and that is when everything went strange. I had to manualy fix duration_correction_factor to fix my completion times, active_frac to get SETI to download WUs, and long_term_debt to get WCG to download. I don't know what I did wrong, but everyone was very helpful to get it fixed. I had previously upgraded two other machines from 6.6.0 to 6.6.20 with no problems at all. I have two more to do but I am not sure I am ready for it yet. - Mitch |
![]() Send message Joined: 29 Aug 05 Posts: 15480 ![]() |
OK, read pre-BOINC 6.6.20 then. 6.6.0 was still using the old method, the new method came into swing in 6.6.6 |
![]() Send message Joined: 29 Aug 05 Posts: 15480 ![]() |
I wonder if it all would have corrected itself given a week or 3 to settle in? I reset one of my two computers at one time and it immediately started learning about how things went the new way. The other machine I did not reset at the time was still out of touch with reality 4 weeks later. I only reset it last week and it's now quickly getting up to speed in the BOINC world. It'll be interesting to see how it will react to my 2 week "vacation", from Friday onwards. ;-) |
Send message Joined: 21 Mar 09 Posts: 20 ![]() |
OK, read pre-BOINC 6.6.20 then. 6.6.0 was still using the old method, the new method came into swing in 6.6.6 Sorry, I was confued by your earlier post of: "BOINC does not reset debts, even though the method of how these debts are used changed between 6.4 and all versions before that and 6.6" Does that mean that it is still likely I will have to take manual actions when I convert my other 6.6.0 machines to 6.6.20. I guess another approach would be to just let BOINC run dry and do a reset. - Mitch |
Send message Joined: 5 Mar 08 Posts: 24 ![]() |
I have just installed 6.6.20 and have had the following problems. 1) BoincDV now comes up with tatally different figures for LTD from what was produced 2 days ago. Then I had a mixture of pluses and minuses adding up to zero. Now I have all negative except for LHC which is zero when it was +240,000. 2) My list of WUs for Seti doesn't match the list they have for me. Some are the same but others are different and recently returned WUs do not appear. Any ideas, please? Mike |
![]() Send message Joined: 8 Jan 06 Posts: 448 ![]() |
I have just installed 6.6.20 and have had the following problems. It's normal. 6.6.x has a new scheduler and LTD has been changed so that now it has a highest value of 0 other projects will be negative. STD still has values + and - |
![]() Send message Joined: 20 Dec 07 Posts: 1069 ![]() |
2) My list of WUs for Seti doesn't match the list they have for me. Some are the same but others are different and recently returned WUs do not appear. The replica database is out of sync. See the Technical News at SETI. Gruß, Gundolf Computer sind nicht alles im Leben. (Kleiner Scherz) ![]() |
Copyright © 2024 University of California.
Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU Free Documentation License,
Version 1.2 or any later version published by the Free Software Foundation.