| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Active Research
Forum: Africa Rainfall Project
Thread: Not enough time given for a task |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 102
|
|
| Author |
|
|
Mike.Gibson
Ace Cruncher England Joined: Aug 23, 2007 Post Count: 12439 Status: Offline Project Badges: 
|
Peter
With WCG you can specify how many of each of their projects you want to run at a time. There is a module called app_config.xml which specifies the maximum number of each sub-project that you want to run, or rather the maximum to run a time. It is always advisable for the total to add up to more than your total threads in case one sub-project runs out of units. It is useful because it is advisable to restrict arp to half your threads and/or mip to one third because of capacity issues. Post here if you want a copy and instructions. Mike |
||
|
|
Mr P Hucker
Advanced Cruncher Joined: Aug 12, 2006 Post Count: 74 Status: Offline Project Badges:
|
Peter With WCG you can specify how many of each of their projects you want to run at a time. There is a module called app_config.xml which specifies the maximum number of each sub-project that you want to run, or rather the maximum to run a time. It is always advisable for the total to add up to more than your total threads in case one sub-project runs out of units. It is useful because it is advisable to restrict arp to half your threads and/or mip to one third because of capacity issues. Post here if you want a copy and instructions. Mike I know all about app_config, I use it a lot for GPU adjustments on other projects. Not sure why it would be better to restrict ARP to less at once? My problem didn't arise from doing lots at once, that would have made no difference. The problem is even doing just one, if it's interrupted by a multicore project like Milkyway or Primegrid or LHC, Boinc leaves the ARP until the last minute as it can't fill all the cores. Something needs to be fixed at either ARP or Boinc itself, because leaving something to the very last minute is going to make things late. It never calculates the deadline correctly, because of a multicore task also running, the user doing something slightly different to average that day, etc. To get round Boinc's stupidity, projects have to allow tasks to come back later than the stated deadline, using a trickleup if you like so you know it's actually being worked on, Primegrid does this and very well. |
||
|
|
tazzduke
Advanced Cruncher Australia Joined: Aug 8, 2009 Post Count: 56 Status: Offline Project Badges:
|
Peter With WCG you can specify how many of each of their projects you want to run at a time. There is a module called app_config.xml which specifies the maximum number of each sub-project that you want to run, or rather the maximum to run a time. It is always advisable for the total to add up to more than your total threads in case one sub-project runs out of units. It is useful because it is advisable to restrict arp to half your threads and/or mip to one third because of capacity issues. Post here if you want a copy and instructions. Mike Greetings Mike Just curious, could we not be pushing people to use their device profiles better, especially for those who have 4 or less computers, where you can stipulate how many of each task you run, especially the ARP and MIP tasks. I found this easier to handle than working with and app_config.xml (I have used these on other projects). Also I note that on other projects, like Primegrid, they have more than 4 device profiles, but you can also specifiy how many tasks and how many cores you can each for these tasks. Just some food for thought. Cheers  |
|||
|
||||
|
Mike.Gibson
Ace Cruncher England Joined: Aug 23, 2007 Post Count: 12439 Status: Offline Project Badges:
|
tazzduke
Which way are you suggesting that we should use? Mike |
||
|
|
Mike.Gibson
Ace Cruncher England Joined: Aug 23, 2007 Post Count: 12439 Status: Offline Project Badges:
|
Peter
The problem that I was referring to with arp and mip is that with most machines, they slow down with higher proportions of threads than I said. If you are crunching other projects as well as WCG, then you have a 2 tier selection issue because WCG counts as a project and arp & mip count as sub-projects, so they only get a share of WCG's portion. Mike |
||
|
|
tazzduke
Advanced Cruncher Australia Joined: Aug 8, 2009 Post Count: 56 Status: Offline Project Badges:
|
Greetings Mike
----------------------------------------Personally I think what WCG has available to use is great, maybe with some extra device profiles though. Regards |
||
|
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1679 Status: Offline Project Badges: 
|
I would like to have some additional device profiles available (without using tricks).
----------------------------------------Cheers, Yves |
||
|
|
Mr P Hucker
Advanced Cruncher Joined: Aug 12, 2006 Post Count: 74 Status: Offline Project Badges:
|
Peter The problem that I was referring to with arp and mip is that with most machines, they slow down with higher proportions of threads than I said. If you are crunching other projects as well as WCG, then you have a 2 tier selection issue because WCG counts as a project and arp & mip count as sub-projects, so they only get a share of WCG's portion. Mike They don't slow considerably. I'd say you get 75% throughput per core when you use hyperthreading. |
||
|
|
Mr P Hucker
Advanced Cruncher Joined: Aug 12, 2006 Post Count: 74 Status: Offline Project Badges:
|
I would like to have some additional device profiles available (without using tricks). Cheers, Yves What are these tricks you speak of? |
||
|
|
KerSamson
Master Cruncher Switzerland Joined: Jan 29, 2007 Post Count: 1679 Status: Offline Project Badges:
|
Editing the link in the address bar and editing the web page content for activating the new profile.
----------------------------------------Yves |
||
|
|
|