Message boards : BOINC client : does a project can abuse the boinc client resources ?
Message board moderation
Author | Message |
---|---|
Send message Joined: 13 Aug 09 Posts: 14 |
Since few weeks I'm observing that one of the 3 project I'm subscribed too sends work units with a rather short deadline. This seems to force my client to always run that project (== 50% of all resources) and to a share the remaining 50% to each of the 2 other projects (== 25 %). I'm running 6.1.242 at under Gentoo Linux and would expect to have that project only running 33 1/3 % of the whole time. |
Send message Joined: 29 Aug 05 Posts: 15581 |
The projects are? The resource shares are? But in essence, yes this is normal. BOINC will try its upmost best to return all work by deadline, and if that means that when one project does a lot of shorties that it takes time away from the other projects, then so be it. However, when that project is done with the shorties, you will see that BOINC will then pay back the extra time it has given to that one project, to the other projects. As long as you don't do any manual interventions (like suspending work or projects). |
Send message Joined: 13 Aug 09 Posts: 14 |
Thx for your explanation. The project is "malariacontrol.net". In the mean while I suspended the project. It uses too much RAM too. |
Send message Joined: 6 Jul 10 Posts: 585 |
Toralf, Out of curiosity, what are the current Malaria.net deadlines, and what are your cache/buffers settings for that host (connect about every/additional buffer or what they're called in the client 7 version minimum work buffer / maximum additional work buffer)? Generally my experience is, that if the sum of those values is at least 1 day less than the shortest deadline of work cached [your malaria, at least you will not see the priority treatment of any project, just over time the share of total time you've specified. The new client behavior is actually kind of to work one project, then when enough is done to work on the next project etc etc. Much less switching nor keeping 1 or more units ready to crunch for all active projects. If a host is doing both GPU and CPU, that will be more complex, but not deviate much from CPU alone. --//-- |
Send message Joined: 13 Aug 09 Posts: 14 |
Currently I'm using 7.0.29, my network usage is 0.1 day (minimum work buffer) 0.1 for Max. additional work buffer. The project starts with 2 hours, then 4 hours, the current task is about 22 hours. As far as I can observed it, the project (reactivated yesterday) was active the whole time, the Boinc and the World Community Grid shares the other 50 %. . I deactivated MalariaControl again, because it eats too much RAM (and the system/CPU is unresponsive too) from my a 2GB RAM of my notebook. |
Copyright © 2025 University of California.
Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU Free Documentation License,
Version 1.2 or any later version published by the Free Software Foundation.