| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: Hmm, what's this??? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 11
|
|
| Author |
|
|
keithhenry
Ace Cruncher Senile old farts of the world ....uh.....uh..... nevermind Joined: Nov 18, 2004 Post Count: 18665 Status: Offline Project Badges: 
|
Stumbled upon what could possibly be an interesting bit of BOINC trivia. I happend to open Task Manager to the Processes tab and noticed two wcg_faah_autodock.... processes. That seemed curious so I looked at BOINC Manager and noticed that I had both the current running WU and a preempted WU with non-zero CPU time. Played a bit with suspending the running WU so a new one would start. Each new WU with a non-zero CPU time resulted in a new wcg_faah_autodock.... or wcg_hpf2_rosetta... process. Now only the one for the active WU had non-zero CPU numbers in Task Manager but it got me to wondering. That has to impact memory some amount. This could make things interesting if you're wanting to do that rare kill a WU from Task Manager (vs aborting it in BOINC Manager). Might be something to avoid (having more than 1 or 2 running/pre-empted/suspended tasks with non-zero CPU time) if you're machine is comstrained???? Doubt that this would make much if any difference to most folks but just maybe.........
---------------------------------------- |
||
|
|
retsof
Former Community Advisor USA Joined: Jul 31, 2005 Post Count: 6824 Status: Offline Project Badges: 
|
Check your BOINC task display. I have seen this with a preempted workunit with due date 7 days from the download date. The currently running workunit will have a due date of 1 day (tomorrow) and is likely one of the #3 workunits quickly needed for a validation. (A previous #0, #1 or #2 workunit could have had an error when uploaded by somebody else.) This one should finish, and the preempted workunit will then continue where it left off.
----------------------------------------If this is not what is going on, please add some more information.
SUPPORT ADVISOR
----------------------------------------Work+GPU i7 8700 12threads School i7 4770 8threads Default+GPU Ryzen 7 3700X 16threads Ryzen 7 3800X 16 threads Ryzen 9 3900X 24threads Home i7 3540M 4threads50% [Edit 2 times, last edit by retsof at Jul 15, 2006 1:24:20 AM] |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Another observation, or maybe not understanding the previous post is that a process will stick in memory, when being suspended, but not having reached its first checkpoint....on HPF2 that might in fact be the first segment, on FAAH, several minutes. A typical event for those 'managing' the BOINC workorder....and yes it will hog the 80/90mb of ram and 277 / 282mb swapfile until its allowed to reach that first verification point
---------------------------------------- The why, i'm not sure, but maybe similar to that period in UD, where there is no simulation of the structure/molecule 
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
The why a result that has not checkpointed is left in memory is simple. There are some projects that never checkpoint, and in order to make progress on the result it has to be left in memory. If a project is removed from memory, it resets back to the last checkpoint (or the beginning if it has never checkpointed).
|
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Thanks John MCLeod VII..., the return to checkpoint is well known here, its that first hump,,,why should the process be in-considerate on any consequent checkpoints, but prior to the first? Suspision is, that in its 1stperiod, its trying to determine if there are checkpoints/segments/chunks at all.....anyway its going to be solved in version 5.6 of BOINC...its going to wait with switching until checkpoint is reached
----------------------------------------
WCG
----------------------------------------Please help to make the Forums an enjoyable experience for All! [Edit 2 times, last edit by Sekerob at Jul 15, 2006 3:24:32 AM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Thanks John MCLeod VII..., the return to checkpoint is well known here, its that first hump,,,why should the process be in-considerate on any consequent checkpoints, but prior to the first? Suspision is, that in its 1stperiod, its trying to determine if there are checkpoints/segments/chunks at all.....anyway its going to be solved in version 5.6 of BOINC...its going to wait with switching until checkpoint is reached With the exception of a project that has a possible deadline miss. If some project has a possible deadline miss then the earliest deadline result from all of these projects will be switched to. BTW, I know very well that this is already implemented and being tested in 5.5 :) |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
John, how is it set in this test-series? knreed wrote its controlled through the website profile. Will the throttle be on a by-project basis or will it be across the board? Been happily running with a throttle-down thru Threadmaster 1.12 which allows control on a by-process basis, long as one tells it what the name is of the sciences.
----------------------------------------
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
keithhenry
Ace Cruncher Senile old farts of the world ....uh.....uh..... nevermind Joined: Nov 18, 2004 Post Count: 18665 Status: Offline Project Badges:
|
I do get the occasional WU with the 2-3 day due date versus the usual seven. Not sure if that was a part of how I came to have a pre-empted WU with 0% progress and non-zero CPU time but it could be. The evidence of that is gone now. It does seem and makes sense that the task would remain in memory until that initial checkpoint as until that first checkpoint, there's nothing yet written to disk. Once that happens, the task is going to start from that checkpoint so there's no reason to keep the task in memory unless you've set your profile that way. Given the long times between checkpoints for HPF2 WU's, if you're only crunching that project, it may possibly be worth considering keeping WUs in memory, at least until we get the new BOINC that won't switch WU's on us until a checkpoint. Not sure how significant that would add up to be but if you want to squeeze every possible bit out of your crunching time, it's a thought.
---------------------------------------- |
||
|
|
keithhenry
Ace Cruncher Senile old farts of the world ....uh.....uh..... nevermind Joined: Nov 18, 2004 Post Count: 18665 Status: Offline Project Badges:
|
Oh, one other thing I've noticed with BOINC 5.4.9 is that the values in the To Competion column seem to be considerably more accurate that before. That means you can tighten up your Connect to Server value. I can't see much if any reason to go beyond 2.0-3.0 for this under normal circumstances. For me, I see this as the best improvement in 5.4.9.
----------------------------------------As for "managing" your WUs, I've seen some discussion about that and I've tried some experimenting. With 5.4.9, I don't see that it really gains you that much. Typically, you will want to crunch the WU that has the closest due date first. Given how 5.4.9 will automatically switch to "older" WUs, you have to leave that WU suspended to prevent that. There you then risk running out of work or crunching on a new WU that you got because you had no WU to run. Besides, how practical can it be to try to managing WUs if you have multiple devices? |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
John, how is it set in this test-series? knreed wrote its controlled through the website profile. Will the throttle be on a by-project basis or will it be across the board? Been happily running with a throttle-down thru Threadmaster 1.12 which allows control on a by-process basis, long as one tells it what the name is of the sciences. Not quite certain that I understand the question. Whether to keep results in memory when they are pre-empted is a Global option, and is controlled through your Global Preferences. In 5.5.x BOINC will always wait for a checkpoint to switch - unless it a project needs extra CPU time. |
||
|
|
|