Can anybody explain how BOINC uses multiple projects especially with regard to downloading WU's.

If you are attached to two projects, for instance, and your settings are for say 3 days worth of work should BOINC download 3 days worth from each project?

If you allocate even resources to each project does BOINC allow both projects to run at once or would it process a work unit from one then a work unit from the other.

Neil

> Can anybody explain how BOINC uses multiple projects especially with regard to
> downloading WU's.
>
> If you are attached to two projects, for instance, and your settings are for
> say 3 days worth of work should BOINC download 3 days worth from each
> project?
>
> If you allocate even resources to each project does BOINC allow both projects
> to run at once or would it process a work unit from one then a work unit from
> the other.

It will fill the queue from 1 project and begin processing those. With an even resource split, when the low water mark gets hit it will refill from the other project. It will only process one wu at a time - whatever is next in the queue.

If the next up project is not available when the low water mark gets hit, it will default back to the other.

I have a 50/50 split on all my hosts between seti and predictor, but I have the amount of work to get set very low (.5 days). With these settings, I get 1 or 2 wu's alternately from each project - but when seti's down, it keeps refilling from predictor.

it doesn't handle it well, I have a 1:100 split pah:sah but my cache still gets filled up with pah ;/ (indeed now to the point where seti has run dry and can't refil).

Some of that might be due to seti downtime tho

> it doesn't handle it well, I have a 1:100 split pah:sah but my cache still
> gets filled up with pah ;/ (indeed now to the point where seti has run dry and
> can't refil).
>
> Some of that might be due to seti downtime tho

The cache will fill completely with work from whichever project should be next based on your split (or what is available if a project is down). If you are set for a large cache and seti is down at the very moment it tries to connect, then it will default to the next project and fill the cache from there.

If you want almost all seti, and only want predictor when seti is not available, then you will have to set your cache very low in combination with your 1:100 split. If you set it to no more than 0.5 days, whenever seti is down and it falls back to predictor the most it will download will be a couple wu's. If you have it set large trying to grab up a big seti cache when units can be had, it also means it's going to grab up a big predictor cache when seti units can't be had.

Does anyone know how BOINC should be prioritising the WU's to crunch?

I posted in another thread that I was only getting Predictor WU's, but overnight my computer downloaded 13 SETI WU's. All are sitting as 'Ready to run'.

I currently have 52 Predictor WU's 'Ready to run' and 13 SETI WU's in the same state. My computer isnt working on any of the SETI WUs, and is currently crunching through the Predictor units - I thought it should be doing two simultaneously and allocating processor resources based on the split I have specified, but obviously that isnt the case... does anyone know what it should be doing in this scenario?

> I currently have 52 Predictor WU's 'Ready to run' and 13 SETI WU's in the same
> state. My computer isnt working on any of the SETI WUs, and is currently
> crunching through the Predictor units - I thought it should be doing two
> simultaneously and allocating processor resources based on the split I have
> specified, but obviously that isnt the case... does anyone know what it should
> be doing in this scenario?

It is operating correctly. It will only work on one project at a time, not simultaneously.

> > it doesn't handle it well, I have a 1:100 split pah:sah but my cache
> still
> > gets filled up with pah ;/ (indeed now to the point where seti has run
> dry and
> > can't refil).
> >
> > Some of that might be due to seti downtime tho
>
> The cache will fill completely with work from whichever project should be next
> based on your split (or what is available if a project is down). If you are
> set for a large cache and seti is down at the very moment it tries to connect,
> then it will default to the next project and fill the cache from there.
>
> If you want almost all seti, and only want predictor when seti is not
> available, then you will have to set your cache very low in combination with
> your 1:100 split. If you set it to no more than 0.5 days, whenever seti is
> down and it falls back to predictor the most it will download will be a couple
> wu's. If you have it set large trying to grab up a big seti cache when units
> can be had, it also means it's going to grab up a big predictor cache when
> seti units can't be had.
>
>
Basically sound advice here. However it is the difference between the high and low mark that is important in limiting the amount of work that you get from the non-prefered project. BOINC fills when it reaches the lower and gets enough work to pass the upper. So 2.5 to 3 will get fewer workunits at a time than 1 to 3 would.

John Keck
BOINCing since 2002/12/08

> Does anyone know how BOINC should be prioritising the WU's to crunch?
>
> I posted in another thread that I was only getting Predictor WU's, but
> overnight my computer downloaded 13 SETI WU's. All are sitting as 'Ready to
> run'.
>
> I currently have 52 Predictor WU's 'Ready to run' and 13 SETI WU's in the same
> state. My computer isnt working on any of the SETI WUs, and is currently
> crunching through the Predictor units - I thought it should be doing two
> simultaneously and allocating processor resources based on the split I have
> specified, but obviously that isnt the case... does anyone know what it should
> be doing in this scenario?
>
BOINC runs only one WU per processor, it does not do fine timeslicing. It currently crunches WUs in the order of the nearest deadline.