| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: Resource share question |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 5
|
|
| Author |
|
|
wolfman1360
Senior Cruncher Canada Joined: Jan 17, 2016 Post Count: 176 Status: Offline Project Badges: 
|
Hello.
----------------------------------------I have WCG set at 100% resource share. However, it is set for subprojects that don't have a steady flow of work, so Rosetta is set for 0. When WCG units are in progress, but all cores are not in use, will Rosetta still take the remaining cores? Or must I increase the resource share to a number above 0 so idle cores are utilized? thanks.
Crunching for the betterment of human kind and the canines who will always be our best friends.
AWOU! |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Yes, a 0 resource project is only asked for work when there are idle resources. Don't know if this is due too app_config restrictions i.e. when there is work ready to start but not allowed to use the resource, i.e. set 7 threads for mip and 1 for fahb, 7 more mip ready to start and no fahb at all. Something for trial and error.
|
||
|
|
Mike.Gibson
Ace Cruncher England Joined: Aug 23, 2007 Post Count: 12425 Status: Offline Project Badges:
|
It could well be that 0 does just that because app_config 0 does not prevent work happening.
However, there are 2 WCG projects where work is readily available - mcm1 and mip1. You could limit their cache to the total threads on the machine in Device Profiles and have a higher setting for the other projects. They are short computation units so the intermittent projects you prefer would kick in very quickly once downloaded. Mike |
||
|
|
wolfman1360
Senior Cruncher Canada Joined: Jan 17, 2016 Post Count: 176 Status: Offline Project Badges:
|
It could well be that 0 does just that because app_config 0 does not prevent work happening. However, there are 2 WCG projects where work is readily available - mcm1 and mip1. You could limit their cache to the total threads on the machine in Device Profiles and have a higher setting for the other projects. They are short computation units so the intermittent projects you prefer would kick in very quickly once downloaded. Mike I was debating that, too - or at least have one project - MCM, maybe - cache all threads minus one - 7 in the cache with 8 threads, for instance, so if there is no FAH2, HSTB, or ARP available Rosetta would then crunch on that last remaining thread, as I do want it to get just a little work. This is my first time really messing with resource share settings - normally I just set and forget. But from what I'm reading if I set Rosetta much higher, since WCG has been crunching this long, Rosetta will take over quite drastically for a while before the boinc scheduler settles down. Thus I might miss out on potential days, if not weeks, of work. I hope that makes sense and what I'm asking isn't too picky or impossible. Just want to make sure I'm understanding this correctly. thanks
Crunching for the betterment of human kind and the canines who will always be our best friends.
AWOU! |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Do the test, set WCG in app_config to <project_max_concurrent>7</project_max_concurrent> with 8 threads and see if it starts to fetch a Rosetta. Of course you would want to restrict Rosetta too in app_config to <project_max_concurrent>1</project_max_concurrent> since one of the rules is, that BOINC will fetch as much as it can get in the request as needed to fill the buffer, at least that is what I seem to remember, not just 1 second of work which will than be 1 task.
----------------------------------------[Edit 1 times, last edit by Former Member at Jan 15, 2020 10:48:14 PM] |
||
|
|
|