| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Completed Research
Forum: Help Cure Muscular Dystrophy - Phase 2 Forum
Thread: Help please |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 12
|
|
| Author |
|
|
falconet
Advanced Cruncher Joined: Jun 25, 2009 Post Count: 50 Status: Offline |
Hi all!
----------------------------------------Since HCMD 2 has the cut off point at 12 hours if not all positions have been processed then a WU with those remaining positions is created right? So I was wondering if it is possible to get only these jobs?Is it? thanks anyway |
|||
|
||||
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
falconet,
----------------------------------------How many do you think there are of these 12 hour cut off jobs? I've yet to see a single one and done 2160 jobs since this morning. Think that's quite a bit of coding to add for a very low number. Consider, if the initial cut off is >60% complete at 6 hours, how often would the last 40% need to go the next full 6 hours? edit: Think if a job does run to the 12 hour cut off, that position underhand could also take you to 13, 14, 15 and more hours.
WCG
----------------------------------------Please help to make the Forums an enjoyable experience for All! [Edit 1 times, last edit by Sekerob at Nov 19, 2009 9:13:56 PM] |
||
|
|
falconet
Advanced Cruncher Joined: Jun 25, 2009 Post Count: 50 Status: Offline |
ok thanks
----------------------------------------i've received many in my laptop and they run for no more than 2 hours.I wanted more |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
The project average run time is 4.52 hours presently, so like me, I'd expect for your computer(s) to also receive a fair spread of long and short jobs.
----------------------------------------
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
falconet
Advanced Cruncher Joined: Jun 25, 2009 Post Count: 50 Status: Offline |
yes i received in the last few days 6 normal WU's and 4 short WU's
---------------------------------------- |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
You can always up the size of your cache if you're worried about running out of work units between times of connecting to the internet.
|
||
|
|
Rickjb
Veteran Cruncher Australia Joined: Sep 17, 2006 Post Count: 666 Status: Offline Project Badges: 
|
You can often increase your proportion of shorter jobs by changing the length of your work cache up and down at particular times. (BOINC client Advanced >> Preferences >> Network Usage >> Additional work buffer)
This can work because the long jobs are usually the parent WUs, which have only 2 small numbers on the end of their names, eg: CMD2_0164-MYH2A.clustersOccur-2OU2_A.clustersOccur_156_0 The descendant WUs are usually shorter, and these have extra pairs of numbers at the end of their names, eg a 2nd-generation WU: CMD2_0164-MYH2A.clustersOccur-2OU2_A.clustersOccur_156_209474_211001_0 3rd-generation WUs (4 extra numbers) and 4th-generation WUs (6 extra numbers) are even shorter on average, as you have noticed. The parent (long) WUs seem to come in batches, and after each batch of these, its descendant WUs are sent out. So, if you get a parent WU, decrease your work buffer setting so you won't get any more work until you have crunched your queue down a bit, by which time some shorter WUs may be coming. Then if you get some shorter-looking WUs, temporarily increase your work buffer to try to get some more. You will still get some parent WUs when you don't want them, but you can alter the odds. Now if everybody sees this and acts on it, there might be WCG server problems. So keep it secret  |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
You can often increase your proportion of shorter jobs by changing the length of your work cache up and down at particular times. (BOINC client Advanced >> Preferences >> Network Usage >> Additional work buffer) hmmm, well, the distributor/shared memory does not know concepts of long and short jobs. It just holds I think 3500 and what ever happens to be in there comes out. Even if your client asks for 1 second of work, it will send the next in queu, whatever size it is, thus only by guesstimating where in the crunching cycle a batch is and manipulating the client buffer... well wish you good luck mm'ing that ;>)
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3716 Status: Offline Project Badges: 
|
hmmm, well, the distributor/shared memory does not know concepts of long and short jobs. Right. So the reason of what Rickjb describes is simply not there. It is more probably in how it is fed. I think that the process which generates WUs below the parent level is run periodically and that the result of this process is not mixed with the parent batches. While you were off there have been times where every WUs were exclusively parents, or children, or grand-children, or even great-grand-children with shorter deadlines, although this last category was more difficult to catch.  While we were crunching the big 147 and 148 batches I remember having been able to load 150 grand-child WUs within minutes without any single intruder.Nowadays it seems that it is either parents, or children and grand-children mixed together. I don't know if the techs have intentionally changed something in the feeding procedures or if it is the natural consequence of - less demanding batches - better peer balancing which have happened since then. I would not be surprised that the techs had changed something because, as you know, alternating such series in an unattended HCMD2-only machine can have serious consequences on the DCF and consequently on the flow of work: at the end of a period of crucnhing grand-children only your machine may be loading a bunch of parents with too-low estimated durations, and when the first parent completes and re-adjusts the DCF this machine will not need any more work during one day or more! |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
This is from a HCMD2 exclusive crunch perspective I guess. I don't, and thus not thanks to HCMD2 see everything that drives the 1.5 day buffer yoyo not so much. Fortunately since the rDCF goes down only slowly, the few shorter don't have too much impact in the multi-project mix.
----------------------------------------Here, filtered out these jobs from the past 3 days ... and wish those that manually tune the buffer size to get small or big good luck. CMD2_ 0160-COPGA._ 4_ 61761_ 63087_ 62785_ 63087_ 1-- 21-11-09 08:30:55 22-11-09 09:04:12 0.36 6.7 / 6.4 CMD2_ 0164-MYH2A._ 1_ 0-- 20-11-09 11:33:43 22-11-09 09:03:57 6.22 115.6 / 111.6 CMD2_ 0164-MYH2A._ 197_ 1-- 20-11-09 03:28:58 22-11-09 09:03:39 6.63 123.2 / 111.3 CMD2_ 0164-MYH2A._ 312_ 0-- 20-11-09 03:28:21 21-11-09 22:16:13 6.79 125.8 / 114.0 CMD2_ 0160-2DVJ_ A._ 28_ 15645_ 15741_ 0-- 19-11-09 23:48:09 21-11-09 19:01:36 4.55 84.1 / 89.2 CMD2_ 0160-COPGA._ 534_ 196806_ 196879_ 1-- 19-11-09 23:47:48 21-11-09 15:55:27 3.72 68.9 / 63.1 CMD2_ 0159-ACTC._ 10_ 31848_ 32397_ 1-- 19-11-09 23:47:27 21-11-09 15:55:27 6.00 111.1 / 98.9 CMD2_ 0164-MYH2A._ 275_ 1-- 19-11-09 19:35:38 21-11-09 11:40:34 7.07 130.8 / 119.4 CMD2_ 0164-MYH2A._ 341_ 1-- 19:35:16 21-11-09 08:31:15 5.86 108.3 / 100.5 CMD2_ 0163-COPGA_ 725_ 217647_ 217723_ 0-- 19-11-09 16:22:23 21-11-09 08:31:15 7.51 138.8 / 125.4 CMD2_ 0163-COPGA_ 43_ 62377_ 62510_ 0-- 19-11-09 16:22:08 20-11-09 17:07:12 3.11 57.5 / 52.9 CMD2_ 0163-MYH2A._ 48_ 120519_ 121484_ 1-- 19-11-09 11:35:22 20-11-09 08:08:19 5.86 108.5 / 98.7 CMD2_ 0163-COPGA._ 1_ 19824_ 21225_ 0-- 19-11-09 06:59:18 19-11-09 23:47:48 2.72 50.3 / 50.7 CMD2_ 0163-MYH2A._ 543_ 398456_ 398751_ 1-- 19-11-09 06:55:46 19-11-09 22:01:06 2.40 44.3 / 39.4 CMD2_ 0163-MYH2A._ 106_ 0-- 18-11-09 08:23:58 19-11-09 06:55:45 6.00 111.1 / 103.7 That the child generation is concentrated is spock logical. There's lots of cron jobs to optimize efficiency and balance the load.
WCG
----------------------------------------Please help to make the Forums an enjoyable experience for All! [Edit 1 times, last edit by Sekerob at Nov 22, 2009 9:59:54 AM] |
||
|
|
|