| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: The Clean Energy Project - Phase 2 Forum
Thread: Incorrect CPU usage time being reported and low credit granted [problem terminated] |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 33
|
|
| Author |
|
|
Ingleside
Veteran Cruncher Norway Joined: Nov 19, 2005 Post Count: 974 Status: Offline Project Badges: 
|
I found project dependent variation between CPU Time and Run Time, Docking 9% difference HCC 28% CEP2 42%  Hmm, while individual results can give so large difference if you're using your computer, if this is the average over many results, something doesn't look right... Now, I don't run Linux, but it can still be interesting to compare with my windows-computer. All is how much longer the run-time is compared to cpu-time, averaged over all results for the various WCG-sub-projects: % - # - project 1.8 - 423 - CMD2 2.7 - 446 - DDDT 2.5 - 620 - E0000 5.5 - 141 - er?? 1.1 - 489 - FLU 1.5 - 261 - FAAH 1.6 - 804 - HFCC 1.9 - 213 - n?/m? 2.3 - 665 - R??? 3.0 - 42 - ts?? 2.1 - 296 - X??? 2.07 - 4752 - WCG, all projects, including beta. Which sub-projects that's hiding behind X-results and R-results and the others with non-informative wu-names I don't remember... Now, if there is a significant difference between projects, or if the variation is only due to variable other usage of computer, I don't know. But, all of them has multiple results there the difference between run-time and cpu-time is less than 0.2%.  "I make so many mistakes. But then just think of all the mistakes I don't make, although I might." |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
The WCG exclusive crunchers will recognize most of those prefixes, X probably easiest connected to the X-Ray Crystallography jobs of HCC and the R of RICE, the TS of DDDT-2 and n?/m probably HPF2... they's not in your list. There was a long time ago an FAQ where I collected those but as time goes and interest is barely above zero, they get forgotten.
----------------------------------------
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
pirogue
Veteran Cruncher USA Joined: Dec 8, 2008 Post Count: 685 Status: Offline Project Badges: 
|
Could you provide the output of 'uname -a' and 'mpstat -P ALL'? I started running HCCs last night, so this may not tell you anything.mpstat -P ALL Linux 2.6.33.5-124.fc13.i686 (fed-x2) 07/08/2010 _i686_ (2 CPU) 08:24:45 AM CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle 08:24:45 AM all 0.27 91.73 5.02 0.10 0.00 0.02 0.00 0.00 2.85 08:24:45 AM 0 0.26 91.91 4.97 0.09 0.00 0.00 0.00 0.00 2.78 08:24:45 AM 1 0.28 91.55 5.08 0.11 0.00 0.04 0.00 0.00 2.93 uname -a Linux fed-x2 2.6.33.5-124.fc13.i686 #1 SMP Fri Jun 11 09:48:40 UTC 2010 i686 i686 i386 GNU/Linux |
||
|
|
sk..
Master Cruncher http://s17.rimg.info/ccb5d62bd3e856cc0d1df9b0ee2f7f6a.gif Joined: Mar 22, 2007 Post Count: 2324 Status: Offline Project Badges: 
|
There is no doubt running several projects is more efficient, and I remember measure this several times and finding that if you get a good combination of projects, the points awarded can increase by more than 10%. I agree this is due in part to reduced bottlenecking. When running one task on a given system rather than another task I also saw variation of up to 10%. This is likely architecture related, with tasks capable of exploiting specific algorithms being faster.
To be getting such huge reduction in CPU time is probably another matter, but it could be the result of accumulation of several smaller issues (network driver slowing tasks down during communication, caching methods, limited operating system, not crunching a variety of tasks, being unlucky in that these CEP2 tasks are not especially optimised for my system and may be competing for resources...). I’m going to continue crunching for CEP2 and HCC on that system until I can move back to XP. |
||
|
|
martin64
Senior Cruncher Germany Joined: May 11, 2009 Post Count: 445 Status: Offline Project Badges: 
|
Could you provide the output of 'uname -a' and 'mpstat -P ALL'? Of course. uname -a Linux silver 2.6.32.6 #8 SMP PREEMPT Thu Jan 28 10:51:16 CET 2010 i686 GNU/Linux mpstat -P ALL Linux 2.6.32.6 (silver) 08.07.2010 _i686_ 18:07:39 CPU %user %nice %sys %iowait %irq %soft %steal %idle intr/s Running 2 CEP2 tasks. And to make it clear: This is about CEP2 *only*, all other projects run Ok. Regards, Martin  |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
fliegenpilz
----------------------------------------Thanks for tipping off on these commands. Loud and clear on e.g. martin64's CPU stats of a 32 bit platform: All, the combined cores and 0/1 the individual cores, 10.5% to user, 53% nice (poor), 14.76% to system activity and iowait being the part I don't like... 9.2% Ran it off my quad that has been up for a few days (it told me first I had to install it, so did), 64 bit: Linux xxxxxxxx 2.6.32-24-generic #38-Ubuntu SMP Mon Jul 5 09:20:59 UTC 2010 x86_64 GNU/Linux It is obvious that what I do on the system is barely taxing, so am glad the fastest part goes to science. About 72% of the time here is used by CEP2 per me latest stats and think to be please with the nice part at 94.4%. edit: Lets not forget in martin64's sample the 12% that the system is really idle. Is this some effect of the client not having run for periods or the client being paused during use?
WCG
----------------------------------------Please help to make the Forums an enjoyable experience for All! [Edit 2 times, last edit by Sekerob at Jul 8, 2010 7:52:58 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
You could also try 'mpstat -P ALL 2 1' which shows just the stats from the last 2 seconds (not since reboot).
Do you also have the effect that 'top' shows in its summary with multi-cpu-display 25% idle but the tasks show 100% (usage is *really* 100%).
but 13506 boinc 39 19 364m 150m 9452 R 100 7.5 41:18.49 wcgrid_cep2_qch Seems to me like a flaw in top. Aggregated display shows: Cpu(s): 0.0%us, 6.5%sy, 85.8%ni, 7.7%id, 0.0%wa, 0.0%hi, 0.0%si, 0.0%st which totally confuses me. |
||
|
|
martin64
Senior Cruncher Germany Joined: May 11, 2009 Post Count: 445 Status: Offline Project Badges:
|
You could also try 'mpstat -P ALL 2 1' which shows just the stats from the last 2 seconds (not since reboot). Ok:
This is pretty consistent, I repeated several times with similar results. So if I have well above 90% "nice" (i.e. CEP2?), why is the CPU time so much different from wallclock time? Regards, Martin |
||
|
|
martin64
Senior Cruncher Germany Joined: May 11, 2009 Post Count: 445 Status: Offline Project Badges:
|
Aah, so "top" brings it to the light:
----------------------------------------3713 root 39 19 343m 106m 9544 R 76 5.2 48:59.39 wcgrid_cep2_qch So obviously having boinc on an NTFS file system (i.e. my normal Windows partition) is not exactly a good idea? But why does this process take so much time, although I don't have any swap or something installed? I guess I simply have to wait for the windows version in order to have a better performance.  [edit] When I suspend boinc, mount.ntfs immediately gets virtually inactive. So the extreme CPU load for that process seems to be clearly caused by CEP2. Maybe CEP2 could be optimised in order not to put such a high load on the disk mount process? Does it permanently reload stuff from the disc rather than loading it into RAM and run from there? [/edit] Regards, Martin [Edit 1 times, last edit by martin64 at Jul 8, 2010 9:36:58 PM] |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
2 seconds is of course in no way representative of what happened over the time you logged the result times. Now you have to clean start and let it run for longer, before checking you know what's running. Presently I'm going let it run it with mpstat -P ALL 7200 1 and not touch the system, since my user stat is a percent spread over several days up time.
----------------------------------------fliegenpielz, yes top shows the percent per core up to 100% as does system monitor. I missed though where you get that first set where the Cpu0/1/2/3 is shown. Idle is I suppose the amount of time 'spare' per core from overall available for the nice processes.
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
|