| 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 |
|
|
Rickjb
Veteran Cruncher Australia Joined: Sep 17, 2006 Post Count: 666 Status: Offline Project Badges: 
|
"That the child generation is concentrated is spock logical. There's lots of cron jobs ... "
----------------------------------------Too much trekkie talk there for me to underconstumble. All I know is that for some days the HCMD2 WUs that my AMD has fetched have been mostly parent ones, but the last 4 WUs are children. So if I wanted to minimise my number of parent WUs, I'd try bumping up my work buffer about right now. [Edit]: Decided to put my mouse where my mouth was, and grabbed some WUs with the AMD by gradually increasing the work buffer by 3 days. 5 out of 5 HCMD2 WUs downloaded are children. Of course, I could have kidnapped more, but that's not my aim. I think I've demonstrated that my idea works. (The AMD also fetched 5 FAAH WUs. That's not many WUs for 3 days on a 2-core machine, but of course that's what happens when the last HCMD2 WU crunched runs 9h55m.) ========== [Edit #2 and #3]: I just realised that falconet may want long jobs, not short ones. In that case, do the reverse. When parent WUs are coming through, temporarily increase your buffer, and when shorter ones start arriving, decrease the buffer. [Edit 4 times, last edit by Rickjb at Nov 24, 2009 3:45:48 AM] |
||
|
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3716 Status: Offline Project Badges: 
|
I just realised that falconet wants long jobs, not short ones. You made me reread his post an umpteenth time, and I am still understanding that he wants the jobs with the remaining positions, which should logically be quite shorter. But sure it would be simpler if he clearly said if he would prefer very short of very long jobs.  Not surprised that you could make a refill of child WUs, it has worked every time I tried it (in a HCMD2 exclusive setup, yes, Sekerob). Rick do you mean you got it working with other projects selected too? Interesting. For a HCMD2-only machine I think it will work until the techs change something to avoid these long series which can unbalance unattended HCMD2 machines (I think we have many in my team). Something simple could be to - put the newly created child WUs in a temp directory together with a bunch of parent WUs - get the list of those files - sort the list on a "meaningless" digit like the first one of the "slice" number (the 53rd digit would be a good candidate) - then use the sorted list to feed the distributor when needed. I made a simulation on 127 WU names of mine and the mix of parents and children was not too bad. |
||
|
|
|