| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: Uncovering Genome Mysteries
Thread: UGM: weak performance on Windows |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 15
|
|
| Author |
|
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1684 Status: Offline Project Badges: 
|
Since a couple of days I can observe a poor performance of UGM work units on two Windows machines (XP Pro SP3 x32 on an Athlon x4, Win7 Pro SP1 x64 on a i7).
----------------------------------------In the both cases, I rebooted the systems assuming that some ghost processes would unnecessary use CPU cycles in the background. Nonetheless the situation remains the same. It happens only with UGM on these both hosts. Other projects (OET1 or MCM) run fine and UGM on Linux hosts runs fine as well. I would be curious to know, if other members do notice similar effects. Cheers, Yves --- Edit: please apologize: I mean BoincTasks (not BoincStats)  ---------------------------------------- [Edit 1 times, last edit by KerSamson at Feb 15, 2015 8:59:05 AM] |
||
|
|
mibere
Advanced Cruncher Joined: Jan 31, 2015 Post Count: 57 Status: Offline |
Boincstats reports a CPU efficiency between 50% and 80% instead of the usual 99.x %. May I ask where I can find this info at Boincstats? |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
As per discussions -and- the last technical news update, there's a high frequency writing issue with UGM. Effectively even if you set Write to Disk at most", there continuous writing every few seconds. A new application is in development, ATM not remembering if there was a first beta on this or not. The goal is to not write to disk unless a checkpoint is taken as all other science apps do.
----------------------------------------As for the poor performance, did tweak some HD settings through W8.1 [The "yes I could" department]. Where before running more than 4 concurrent on an octo would cause rapid decline of efficiency, now can run 8 and still have ~99%. A small sample: World Community Grid 7.23 ugm1 ugm1_ugm1_08106_0504_0 04:51:46 (04:48:09) 2/11/2015 11:01:22 AM 2/11/2015 11:03:38 AM 98.76 World Community Grid 7.23 ugm1 ugm1_ugm1_08109_0181_1 03:13:06 (03:10:12) 2/11/2015 10:55:22 AM 2/11/2015 10:56:01 AM 98.50 World Community Grid 7.23 ugm1 ugm1_ugm1_08106_0369_1 04:52:49 (04:49:04) 2/11/2015 10:41:03 AM 2/11/2015 10:41:57 AM 98.72 World Community Grid 7.23 ugm1 ugm1_ugm1_08102_0307_0 04:49:29 (04:46:45) 2/11/2015 9:55:05 AM 2/11/2015 9:56:56 AM 99.06 World Community Grid 7.23 ugm1 ugm1_ugm1_08098_0054_1 04:50:00 (04:47:24) 2/11/2015 9:10:59 AM 2/11/2015 9:12:53 AM 99.10 World Community Grid 7.23 ugm1 ugm1_ugm1_08103_0050_1 03:10:36 (03:08:59) 2/11/2015 8:54:09 AM 2/11/2015 8:57:07 AM 99.15 Of course, no one should be forced to tweak hardware to accommodate set and forget volunteer computing. 99% is the "out of the box" reference expectation [except for CEP2]. Is it BOINCStats or is it WUProp of which I know they keep detailed project stats [for those participating]. [Edit 2 times, last edit by Former Member at Feb 14, 2015 4:09:38 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
The news excerpt from Jan.16:
"The Uncovering Genome Mysteries project has an issue where it writes to the disk very frequently. At this time we are alpha testing some changes that will prevent this from happening. The current version in alpha does have a known issue and is being investigated, as it was causing invalid errors. Once we have a suitable version of the application, we will promote it to beta testing to make sure the changes are valid. " |
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7846 Status: Offline Project Badges:
|
I have noticed a somewhat different sideefeect from UGM that I have not experienced with any other project. On a q6600 using Windows Vista, I get uncontrolled growth of the system files used for backup. Periodicall I need to go and manually have disk cleanup remove the old ones. It takes about 2 weeks to use about 20 to 25 GB of space. I wonder if this could be a sideeffect of the diskwriting phenomenon ? I have not seen this behavior with any of the othher projects. I had MCM1 running on this system for months and this never occurred. It is not a big problem, just a little annoying.
----------------------------------------Cheers
Sgt. Joe
*Minnesota Crunchers* |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
"uncontrolled growth of the system files". What files exactly are so monstrously growing? Certainly I could not be bothered to back up anything in the BOINC data directory and that is the usual 5GB for a 8 core system with a 1 day cache size. System limit set to 10GB for BOINC, so would know if things go big.
|
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7846 Status: Offline Project Badges:
|
"uncontrolled growth of the system files". What files exactly are so monstrously growing? Certainly I could not be bothered to back up anything in the BOINC data directory and that is the usual 5GB for a 8 core system with a 1 day cache size. System limit set to 10GB for BOINC, so would know if things go big. I should have been more specific. "In some editions of Windows Vista the disk might include file shadow copies and older Windows Complete PC backup images as part of restore points. " I clean up this space by deleting all but the most recent restore point. It is these restore point files which are doing the growing. Cheers Edit: spelling
Sgt. Joe
----------------------------------------*Minnesota Crunchers* [Edit 1 times, last edit by Sgt.Joe at Feb 14, 2015 7:16:53 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Maybe it's W7 or W8, but seem to remember you can set how much of a space restore points are allowed to take. This could contain the pointers: https://www.winhelp.us/configure-system-restore-in-windows.html
Rather quaint for UGM to make these grow exorbitantly. |
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7846 Status: Offline Project Badges:
|
It may just be something peculiar to my system. When I get a chance I will poke around a bit to see about setting the size of the shadow file space. I looked for the slider which XP had, but that is gone.
----------------------------------------Cheers
Sgt. Joe
*Minnesota Crunchers* |
||
|
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1684 Status: Offline Project Badges:
|
@SekeRob
----------------------------------------I thank you for mentioning a possible cause by disk writing. Since I did not experience any troubles with disk writing in the past, even with CEP/CEP2, I did not think about this possible cause. However I am surprised to observe on the same host (e.g. i7: 8 threads), that some UGM WUs are computed efficiently and at the same time some others have poor performance. My first assumption would be that too frequent disk writing would impact negatively every computed WUs at this time, not only some of them. I will try to fine tune a little bit in this area; advices are welcome. Cheers, Yves |
||
|
|
|