Originally posted by: Marrkks
first of all, in the eistein @home forums there was mention of the throttle cpu funtion not playing nice.(advanced->preferences->processor usage->use at most xxx%). Is this a known/ common problem or just an einstien@home issue?
Not sure with v6, but atleast with earlier BOINC-clients, it's using a 1-second on/off-cycle. Meaning, if example 90%, you'll use cpu 100% for 9 seconds, for so being off 1 second, and similar for other setups. So, appart for 50% that would give a 1 second on, one second off-schedule, it's too erratic to really be much helpful.
second, does the boinc manager do a good job of balancing multiple projects? example if I have 2 projects running 100/100 does the manager actually give each project 50% cpu time. I have attached to 13projects, 3 don't have work, 4 I have set to "no new tasks", and the rest are crunching away. Can I expect the manager to properly balance the work load as I have set it to?
most likely more questions later cheers :beer:
2 projects both getting 50% of resources should use roughly 50% on both projects, as long as one of the project isn't constantly down that is... It's possible with a large cache-size you'll basically run only one of them for a few days or something, but it should normally average-out.
With very one-sided setups, like one project on 99.99% and another on 0.01%, you'll not get such a good fit, especially if you also "needs" a 10-day cache. This, since if your "main" project for any reason can't keep the cache full, you'll download work from the "backup"-project, and, due to the very low resource-share, it's a good chance this will immediately run in "high priority"-mode... Even if servers is up, if for some reason a scheduler-request doesn't completely fill your cache-size, there's always a few seconds delay before same project contactable again... ample time for BOINC to connect another project instead to keep cache full...
For many projects with roughly equal share, you can also get a fairly good fit, but any projects that is constantly down or out of work won't keep-up with resource-share-settings. Short-term it likely won't follow your resourse-shares, but averaged over example a month it shouldn't be too far from your settings... The larger the cache-setting and the slower the computer, the longer away from your resource-shares you'll likely to land...
While you won't normally miss deadlines, a good rule-of-thumb that was needed back in v4-days is, to set max cache-size below "shortest deadline" / # projects, so with example 10 projects and one project with 5-days deadlines, the cache-size should be below 0.5 days. If you've got a permanent connection and is sharing many projects roughly equally, it shouldn't really be any reason to use anything larger than 0.1 days cache-size.
Facebook
Twitter