| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: Help Conquer Cancer
Thread: HCC 5.07 and page faults |
| No member browsing this thread |
|
Thread Status: Locked Total posts in this thread: 16
|
|
| Author |
|
|
brent1023
Cruncher Joined: Apr 27, 2007 Post Count: 3 Status: Offline Project Badges: 
|
Hi
I just noticed when looking at the task manager (CTRL-ALT-DEL) that this project is producing a very large number of page faults. After about 3 hours on my PC it has 207,000,000 page faults. The Fight Aids run which has gone almost as long as only 1,622,000 page faults. This is having an effect on response to my browser - things are running more slowly. Usually there is no effect on browser or other program response. Is this a problem with the application? With my PC (XP media centre, AMD dual core). Brent |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Hi,
----------------------------------------This 'higher than other projects' PF generation has been flagged to the programmers and it's related performance impact for some. Please see this thread: http://www.worldcommunitygrid.org/forums/wcg/viewthread?thread=16902#135140 cheers
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
brent1023
Cruncher Joined: Apr 27, 2007 Post Count: 3 Status: Offline Project Badges:
|
It is too bad they have not found a way around this in two years. I watched this task in the Boinc manager for a while today. As it continued to tick up on the CPU time column, it also continued to tick UP in the time to completion column.
The page fault numbers I am seeing are way beyond what was happening in the reports back in 2005. Again with comparable CPU time the HCC task has 320 times as many page faults as the Fight Aids task. Response time on my machine is slowed enough that I find it disconcerting. I have removed HCC from my device profile - there are fortunately enough other projects to keep my PCs going. |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
HCC jobs run on a brand new science application completely different from AutoDock which is very very mature.
----------------------------------------Only during a large scale public run do these 'user experience' problems come completely to surface. I see them millions PFs too, but do not seem in to get impacted from my user perspective.... very very fast internet line and net subsystems may be part of this? At any rate, as said the programmers were notified of the impairment and expect them to work on diagnosing this and working towards a solution.... we usually see a few iterative updates in first few months of a new project launch.
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7693 Status: Offline Project Badges: 
|
I may be wrong, but I think those page faults are slowing the process down. I have a 1.6 ghz P4 which took 34 hours to crunch a cancer wu. I checked it periodically and it was running about 600,000,000 page faults per ten hours of cpu time. Somehow I don't think this is aiding the system to crunch efficiently.
----------------------------------------The credit granted was less than 50% of the claimed so I know it is not efficient for this application. That machine runs mostly hpf2 wu's which normally get granted about 90%+ of what is claimed. I will try another one to see how that goes, but I think it would be a better use of resources if I disable hcc on this machine until a fix becomes available. Hope this helps. 1.6 ghz Intel p4 Win 2000 pro 256 mb memory Cheers
Sgt. Joe
*Minnesota Crunchers* |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
I have a dark brown suspicion that may be the cause of the significant variation in run times within a quorum. All that my quad did of HCC were in the 4-4.5 hour range but not seen any quorum yet. Why would these page fault cause for one to slowdown and not the other? If it effected both, the quorums would have agreeing claims.
----------------------------------------
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Sgt.Joe, the profiling I did confirmed that 10-20% of CPU time is wasted by this. If you are severely memory-constrained, then you might have a bigger problem.
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Cannot they set any good programmers on the job.
It looks like those biology scientists never learned to program well. Cannot they give the programming to some good programmers? If they set some good programmers on the job that also understand how a cpu and hardware is working I think they can make crunching 5 times faster or even more. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
HCC 5.15: In task manager I see PF Delta about 60000-70000.
Configuration: Athlon X2 3800+, 2GB dual channel, about 1.5GB free, 1 thread running Also I check this on Athlon 2500+(single core): PF Delta about 600-700. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello s060319,
Cannot they set any good programmers on the job. Page faults are usually caused by 'locality of reference'. If the algorithm does not reference nearby array values, then the cache will not have the requested information. This is probably a necessity of the algorithm, in which case the programmers cannot do anything about it. Lawrence |
||
|
|
|