| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Active Research
Forum: Smash Childhood Cancer
Thread: SCC WU Length |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 25
|
|
| Author |
|
|
Yavanius
Senior Cruncher Antarctica Joined: Jan 21, 2015 Post Count: 191 Status: Offline Project Badges: 
|
I've noticed that the WUs have a little more doubled in run length. I also noticed when I first got the WUs that I got a beta SCC WU which was roughly 8 hours on my Latitude with a 2nd Gen i5 in it.
I'm curious: are we going to be seeing something in the way of 'short' and 'long' WUs or is this because of the different approaches? Cheers, Yavanius |
|||
|
||||
|
TPCBF
Master Cruncher USA Joined: Jan 2, 2011 Post Count: 1932 Status: Offline Project Badges: 
|
I've noticed that the WUs have a little more doubled in run length. I also noticed when I first got the WUs that I got a beta SCC WU which was roughly 8 hours on my Latitude with a 2nd Gen i5 in it. It seems so.I'm curious: are we going to be seeing something in the way of 'short' and 'long' WUs or is this because of the different approaches? But what really worries me are WUs like SCC1_ 0000002_ Bct-A_ 48823_ 0--, which shows a runtime of 1d10h+ now, currently at 99.994% done, but when looking at the properties in the BOINCManager, it just shows "---" for "CPU Time" and "CPU Time at last checkpoint". I have seen this happen before with the newer batch of FAH WUs as well and at times with some ZIKA WUs as well.  Ralf  |
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7581 Status: Offline Project Badges:
|
I've noticed that the WUs have a little more doubled in run length. I also noticed when I first got the WUs that I got a beta SCC WU which was roughly 8 hours on my Latitude with a 2nd Gen i5 in it. It seems so. But what really worries me are WUs like SCC1_ 0000002_ Bct-A_ 48823_ 0--, which shows a runtime of 1d10h+ now, currently at 99.994% done, but when looking at the properties in the BOINCManager, it just shows "---" for "CPU Time" and "CPU Time at last checkpoint".I have seen this happen before with the newer batch of FAH WUs as well and at times with some ZIKA WUs as well. I'm curious: are we going to be seeing something in the way of 'short' and 'long' WUs or is this because of the different approaches? Ralf I have run several hundred of these units and have not seen anything longer than about 3.5 hours. (My hardware is nowhere near an I5). What I think maybe happening to some of these as you describe is that they get hung in the system. The solution for some is to simply reboot. I know there was a problem a while back where they packed too many tasks in each work unit on one of the VINA projects and the run times really ballooned, but I thought they had resolved that issue. Cheers
Sgt. Joe
----------------------------------------*Minnesota Crunchers* [Edit 1 times, last edit by Sgt.Joe at Jan 31, 2017 5:34:11 PM] |
||
|
|
TPCBF
Master Cruncher USA Joined: Jan 2, 2011 Post Count: 1932 Status: Offline Project Badges:
|
I have run several hundred of these units and have not seen anything longer than about 3.5 hours. (My hardware is nowhere near an I5). What I think maybe happening to some of these as you describe is that they get hung in the system. The solution for some is to simply reboot. I know there was a problem a while back where they packed too many tasks in each work unit on one of the VINA projects and the run times really ballooned, but I thought they had resolved that issue. I have run a few dozen WUs on the same machine (and at the same time) as well, with no ill effects on those. Just checked and the recent validated WUs on this machine are between 1.87h and 4.45h CPU time, with PVa jailed ones of up to 4.54h. And unfortunately, rebooting isn't always an option... Cheers  Ralf PS: Btw, this is an i3 with 8GB of RAM, running Windows 8.1 |
||
|
|
uplinger
Former World Community Grid Tech Joined: May 23, 2005 Post Count: 3952 Status: Offline Project Badges: 
|
The workunits for SCC1 are one target and one ligand. This is because they are flexible doing a more exhaustive search on those. They are also flexible work units which cause them to be more difficult. With this, we are not able to size these work units. They do vary largely based on the size of the ligands being compared to with the target.
# Torsions ATOMS cpuavg count T 0 A 9 C 2506.99 X 13 T 0 A 10 C 2046.63 X 38 T 0 A 11 C 2408.69 X 102 T 0 A 12 C 2281.11 X 100 T 0 A 13 C 2604.74 X 58 T 0 A 14 C 3041.75 X 59 T 11 A 28 C 8901.97 X 16 T 11 A 29 C 17356.8 X 15 T 11 A 30 C 8760.19 X 35 T 11 A 31 C 6701.41 X 41 T 11 A 32 C 8246.24 X 46 T 11 A 33 C 7057.73 X 52 T 11 A 34 C 9502.77 X 57 T 11 A 35 C 7057.25 X 52 T 11 A 36 C 12085.7 X 44 T 11 A 37 C 8806.43 X 50 T 11 A 38 C 11517.6 X 48 T 11 A 39 C 10610.2 X 35 T 11 A 40 C 16560.3 X 21 T 11 A 41 C 8731.05 X 27 As you can see, when torsions and atoms get larger in the ligands the runtime increases. Unfortunately we can not size these. Thanks, -Uplinger |
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7581 Status: Offline Project Badges:
|
Uplinger,
----------------------------------------Thanks for the explanation. Cheers
Sgt. Joe
*Minnesota Crunchers* |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Work unit hangups have been an issue with VINA for quite sometime (years). I wouldn't describe it as a common occurrence but it does happen. It seemed to happen more often on certain processor models. When a WU hangs up it isn't always necessary to reboot the machine ( I understand that is the default diagnostic approach on Windows). I have been able to restart the work unit by simply stopping and restarting the BOINC client.
|
||
|
|
Yavanius
Senior Cruncher Antarctica Joined: Jan 21, 2015 Post Count: 191 Status: Offline Project Badges:
|
As you can see, when torsions and atoms get larger in the ligands the runtime increases. Unfortunately we can not size these. Oh, you might misunderstand me. That wasn't a complaint or a concern. I was just merely curious what the difference was with the increased lengths. :) Are the credits going to scale accordingly along with the increased times? Not that we all don't do these all out of the pure intent of heating up our homes somewhat during the winter. ;) That reminds me. Out of curiosity, I was looking at my results and noticed that sometimes I get less credit than claimed and sometimes more. Would explain that or point me to the appropriate post that explains this -/+ variation? ~Y |
||
|
|
TPCBF
Master Cruncher USA Joined: Jan 2, 2011 Post Count: 1932 Status: Offline Project Badges:
|
Work unit hangups have been an issue with VINA for quite sometime (years). I wouldn't describe it as a common occurrence but it does happen. It seemed to happen more often on certain processor models. When a WU hangs up it isn't always necessary to reboot the machine ( I understand that is the default diagnostic approach on Windows). I have been able to restart the work unit by simply stopping and restarting the BOINC client. Stopping and starting didn't do a thing, that was the first I tried, but I was able to reboot the system earlier today and the WU in question started over from scratch, currently in the high 40% range. Just too bad that it wasted one CPU slot for two days... :-(Ralf |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
The operative words "isn't always" is the key. I have even been able to "dislodge" the stuck WU by simply suspend and resume at times but doesn't always work. Don't know if you checked, but after the client was stopped, were there still processes running? If there were still VINA processes running after the client ended (they weren't cleaned up by the parent process) you might have been able to kill those processes then restart the client. Each environment is a little different but it seems better to start with the least drastic action first. I have never understood how casually some support people take rebooting.
|
||
|
|
|