Back Online
log in

Advanced search

Message boards : News : Back Online

Author Message
Kevin
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 4 Feb 12
Posts: 147
Credit: 2,427,829
RAC: 0
Message 162 - Posted: 24 Aug 2012, 15:40:41 UTC
Last modified: 24 Aug 2012, 15:42:31 UTC

Hi Guys,
We're back online now. ~1.5 million dockings ready to go.

We still have some improvements to make and we will be making these over the coming weeks.

Thanks for your help!
Kevin

dziubas
Send message
Joined: 21 Jul 12
Posts: 2
Credit: 510,065
RAC: 0
Message 165 - Posted: 24 Aug 2012, 16:40:43 UTC - in response to Message 162.

somehow i have got over 3000 WUs for one of my pc which has 0.5/0.25 days buffer set (?)

Chris Granger
Send message
Joined: 19 Jul 12
Posts: 21
Credit: 726,988
RAC: 0
Message 166 - Posted: 24 Aug 2012, 22:40:28 UTC

Yeah, the latest batch is sending out work units like mad. They're short, though, so I think I can blast through them well before the deadline is up. My cache is set to 0.05 days and I still got hundreds of WUs. Something is amiss.

Conan
Avatar
Send message
Joined: 20 Jul 12
Posts: 31
Credit: 451,033
RAC: 0
Message 168 - Posted: 24 Aug 2012, 23:30:34 UTC

Yet I can only get one or two at a time, one machine has yet to get any, yet it keeps asking. So I have the opposite problem.
Still it is working, I am getting some work and the project is going great, so all is well.

Conan
____________

Chris Granger
Send message
Joined: 19 Jul 12
Posts: 21
Credit: 726,988
RAC: 0
Message 169 - Posted: 25 Aug 2012, 1:15:35 UTC

You can have some of mine, Conan. I had to set FM@H to No New Tasks to stem the flow for a bit. lol

dziubas
Send message
Joined: 21 Jul 12
Posts: 2
Credit: 510,065
RAC: 0
Message 171 - Posted: 25 Aug 2012, 3:55:32 UTC

it's ok. first hundreds WUs took 20-40 min, so my rough estimation was about 5 days for all, but remaining ones are ~10 times shorter.

Kevin
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 4 Feb 12
Posts: 147
Credit: 2,427,829
RAC: 0
Message 173 - Posted: 25 Aug 2012, 9:14:37 UTC

Yep, there's certainly something wrong here. I'll be fixing it over the next few hours.
Cheers,
Kevin

Conan
Avatar
Send message
Joined: 20 Jul 12
Posts: 31
Credit: 451,033
RAC: 0
Message 174 - Posted: 25 Aug 2012, 11:39:34 UTC - in response to Message 169.

You can have some of mine, Conan. I had to set FM@H to No New Tasks to stem the flow for a bit. lol


Thanks Chris,
However one of my hosts (a 4 core machine) decided it was a bit hungry and gobbled up over 500 work units.
It is chewing through them at a good rate of knots so I think it will get them all completed.
My other 3 computers have about 8 between them.
The joy of BOINC, hardly ever a dull moment.

Conan
____________

Kevin
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 4 Feb 12
Posts: 147
Credit: 2,427,829
RAC: 0
Message 175 - Posted: 25 Aug 2012, 15:55:55 UTC

Ok, we have now set limits on the number of tasks per core. We did this before but somehow this was changed when we transferred to the new server. If you are working on a large number of tasks it will be a while before you are sent new ones.
Thanks for the feedback,
Kevin

Profile John Hunt
Avatar
Send message
Joined: 14 Aug 12
Posts: 1
Credit: 22,268
RAC: 0
Message 176 - Posted: 25 Aug 2012, 16:25:34 UTC

Great work, Kevin !

(wish that all project developers were as quick to respond)

____________

skgiven
Avatar
Send message
Joined: 27 Jul 12
Posts: 19
Credit: 490,044
RAC: 52
Message 198 - Posted: 1 Sep 2012, 12:45:05 UTC - in response to Message 176.

What's the value for the maximum number of tasks to be allocated per core?

I'm seeing lots of ~80s and ~30sec tasks, but some are as low as 24s.

I use a fairly low cache; 0.25days, with a max. add. work buffer of 0.1 (2.4h). Hopefully I'm not downloading too often, or reporting too frequently with these settings. Even though the cache is low I've still experienced a few Boinc can't connect to client glitches over the last few days. I guess these could be caused by other projects, but I think suggestions have been made that this issue is related to high task numbers in cache. I have about 300 in total. I'll change my GPU project away from POEM for a while to see if that helps, but I think MW might be more challenging.
____________

Henk Haneveld
Send message
Joined: 25 Jul 12
Posts: 18
Credit: 78,778
RAC: 0
Message 199 - Posted: 1 Sep 2012, 13:27:36 UTC

As far as I can tell 50 per core.

skgiven
Avatar
Send message
Joined: 27 Jul 12
Posts: 19
Credit: 490,044
RAC: 52
Message 200 - Posted: 1 Sep 2012, 14:21:34 UTC - in response to Message 199.

Thanks Henk.
For me this means a cache setting over 0.05 is irrelevant; even with a full cache of 50tasks per core the 80s tasks would only last 67min, and the 24s tasks would finish in 20min. If there's an outage I'm also doing some RNA work.

skgiven
Avatar
Send message
Joined: 27 Jul 12
Posts: 19
Credit: 490,044
RAC: 52
Message 207 - Posted: 4 Sep 2012, 19:45:31 UTC - in response to Message 200.

Just checked in on a system and watched a bunch of 15sec tasks complete.
It's a bit eerie watching a task complete every 2 seconds.

I might have asked for this already, but is there any chance of creating batches of tasks and sending those out? I would much rather have 20min to 2h tasks than 15sec tasks, and I'm happy enough to run 8h tasks, so long as the failure rates are low.
____________

Kevin
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 4 Feb 12
Posts: 147
Credit: 2,427,829
RAC: 0
Message 208 - Posted: 6 Sep 2012, 10:03:21 UTC - in response to Message 207.
Last modified: 6 Sep 2012, 12:40:48 UTC

Hi skgiven,
For the moment we won't be releasing significantly longer workunits. Once we have finished the short WUs we will move on to the longer ones (those where the ligand has a large number of rotatable bonds). I can't give you a timeframe for this but we are moving through the short workunits quickly.
Regards,
Kevin

Message boards : News : Back Online


Main page · Your account · Message boards


Copyright © 2017 Dr Anthony Chubb