BOINC Priorities
12 Feb 2009 18:30:26 UTC
Topic 194189
(moderation:
I was wondering the other day why BOINC does not finish the WU it has already started on the project is working on, rather than starting a new WU and crunching on that first. Should not BOINC give priority to those WUs it has already started first? I would think so; after all, it would get older the work out the door quicker, and more importantly, out of the cache. It is certainly not a big deal, but I would think it would help on the margins.
I have also been wondering for some time as well if the devs could write a simple right-click to reconfigure the interface under each of the tabs in BOINC manager. I am talking about the project, application name, team, etc. under each tab. I would like to delete some of those catagories, and expand others to my liking. I think something like this would be well recieved, since we all have our own preferences about what information is displayed under each tab. Is this possible?
(Click for detailed stats)
BOINC Priorities
)
Simply put, the CC work scheduler is designed to accommodate the wide range of projects, settings, and preferences possible, while at the same time avoiding missing any deadlines if at all possible.
For the situation you mention two possible reasons come to mind:
1.) The newer task has a shorter net time to deadline than the current work onboard, therefore is seen to be under time pressure right from the start.
2.) One of the conditions where the CC will run a full scheduling simulation is when new work has been fetched. This also results in debt being recalculated and updated. Many times this will result in 'short pitting' one or more tasks which are currently running and resuming others and/or starting new ones depending on the outcome of the STD recalculation.
HTH,
Alinator
RE: I have also been
)
Post wishes to Whish List or BOINC message board...
Michael
Team Linux Users Everywhere
RE: RE: I have also been
)
I took the liberty of posting my queries to the BOINC message baords in this thread if any wish to follow the discussion there.
(Click for detailed stats)
RE: I was wondering the
)
This is / was a topic of discussion on the Dev list also ... though the reception to the idea of giving the participants the ability to have a setting that they could use to cause the CPU scheduler to run tasks to completion on a project and possibly task level was not well received.
The opinion of Dr. Anderson was this was "micromanaging" which is bad ... of course allowing us to micromanage by setting tasks to suspend or abort is somehow Ok ...
Anyway, this issue came up in the context of Soduko which had some tasks outstanding for the 12 series and they wanted them back as soon as possible. The same obtains on LHC@Home ... sadly ... it is not likely that we will see this soon, if ever ... because of the dreaded "micromanagement" label ... Even sadder is the fact that with 4 and 8 CPU systems becoming more and more common this does not restrict the system as much as they imagine ...
But, the work around is to set the switch interval to, say, 720 minutes (as I have) sort of handles this ... though I did have a task that was 100% done and was in "waiting to run" ... anyway, this micromanagement is ok somehow ... I never can figure out the logic of why some things are Ok and others that get to the same goal are not ... though using the switch interval so high screws up the internal calculations a little bit and you get artifacts where tasks will go into high priority mode because of it ... sigh ...
I have concluded (after
)
I have concluded (after several examples) that Einstein is periodically consuming almost 90% of by CPU WHILE I"M WORKING - according to Task Manager. This in spite of the fact that the total allocated for Einstein and SETI is 80%. I can't get any work done when this happens. I don't care if this is an Einstein problem or a Windows Vista problem. I have decided to discontinue working with Einstein for this reason.
Please tell me how to dis-engage.
Thanks - Don
RE: I have concluded (after
)
There is one thing you could try before you abandon EAH, which would be to tell BOINC to not run the science apps while you are using the machine.
However, if you don't want to do that, then the courtesy way to disengage is to detach the host from the project so the project knows you are dropping out and reassigns it. Alternatively you can NNT EAH and abort the tasks manually, then do an project update and suspend the project.
Alinator
RE: I have concluded (after
)
In addition to the advice given above, another idea you can try if you want to effectivlely limit the work that is done for BOINC, I think a good idea would be to limit BOINC to use only one of the two (virtual) cores on your Pentium 4. BOINC will regard your hyperthreading P4 as a dual core system and will start two processes (at lowest priority). This is OK if you want maximum BOINC performance. But if you feel you want less BOINC going on in the background, just set the max number of CPUs to use in multi-processor systems to 1 in the web preferences. BOINC will never use more than 50% of your CPU then, so to speak.
If your concern is about an "unfair" distribution of work between SETI and Einstein@Home, the story is a complex one.
There is no mechanism in BOINC that will allow you to allocate exactly xx% of the CPU to one project ALL OF THE TIME. On a (virtual) dual-core like yours, BOINC will start two app instances at most, so you have either 100%-x or 50%-x/2 (where x is the percentage used for non-boinc tasks) depending on whether the two tasks are from the same project or not.
There is a CPU-throttling feature in BOINC, however, that allows you to set a maximum for the CPU share used by BOINC (you can't set it differently for different projects tho, the most recent setting wins). There are two things to notice about this feature:
First, it works by interrupting the science app at regular intervals to get the right CPU share on average. So what you will see in the task manager is that the BOINC apps will get close to 100% (on an otherwise idle system), in one second, then sleep for some time, then get 100% again. It's not like task manager showing (say) 80% all the time.
Secondly, this feature seems to be buggy in BOINC at the moment and under Windows generates lots of problems with the Eeinstein app, so it's recommended not to use the CPU throttling by BOINC. Restricting the number of cores to use , as described above, is much more effective to limit BOINC activity on your system.
Happy crunching
Bikeman
Bikeman
RE: I have concluded (after
)
If you are still here you could always go into the Boinc settings by double clicking the icon and the go into Advanced, Preferences, network usage and down in the bottom third of that page you will see where you can limit Boinc's running to certain hours of the day. That way you can tell it to only run, for instance, from 8pm to 5am on weekdays and then 24 hours on weekends. That is good for people that feel the cpu is not responsive enough during normal working hours. Be aware though that by clicking the ok at the bottom you are overriding any webpage settings you might have. So if you have it set to 100% under the processor usage tab and only 50% of the cpu's o nthe webpage, it will be overwritten. BE CAREFUL and check BEFORE you click ok!!! Cancel does not save anything and your old settings remain. Their is a button there you can click to revert to the webpage settings if you decide you don't like the local settings.
RE: Advanced, Preferences,
)
I'd use "processor usage", but ok. ;-)
RE: This is / was a topic
)
My idea is similar in that BOINC itself would prioritize the waiting work units before starting new ones. I generally agree with David Anderson on the issue of micromanagement, although I must admit that the lines are admittedly blurry at best. I am unwary as he is of buttons to accomplish the same result, but that is for others to debate.
(Click for detailed stats)