> From: Paul D. Buck <[email protected]>
> To: TarotApprentice <[email protected]>
> Cc: BOINC dev <[email protected]>
> Sent: Sunday, 26 April, 2009 11:13:20 AM
> Subject: Re: [boinc_dev] 6.6.20 and work scheduling
>> I'd suggest:
>> a) Download finishes (more work has arrived - we need to check deadlines)
>> b) Task completes (we have a free resource)
>> or if none of the above every 5 minutes
I'd have to add to that list suspend/unsuspend task
> The reason I keep questioning the download work check deadlines is that why
> might it be fatal if we
> don't check for even as long as an hour?
>
> This is the part I really don't get, what is changing so fast we have to
> check every 60 seconds. I just
> cannot see it. At TSI (default 60 minutes) or tasks end, sure, no problem ...
>
> Other than that, I am at a loss as to why one hour one way or another, one
> download or another, is so
> critical to make sudden changes.
Sometimes tasks take longer than expected. That may then put other tasks into
deadline pressure. Sure I agree there is no reason to check every 60 seconds. I
think 5 mins is enough to allow for this scenerio assuming some other event
didn't already trigger the checking.
> If work fetch is working correctly, we should not be that overburdened. If
> that is what we are
> protecting against, then the Work Fetch is broken because we are getting too
>much work in an
> unsustainable cycle.
Its not just to protect against that, which I agree should work in the 1st
place. Think of the situation with GPUgrid, where their work has short
deadlines (5 days) and therefore may need to run straight away. Maybe some
project gives you a resend with a 24 hour deadline, something needs to get
swapped out in order to meet the deadline.
> This is part and parcel why I was arguing, unsuccessfully, for an end-to-end
> re-thinking of what we
> are doing back a couple months ago when discussing the approaches for GPUs
>... because we are
> going to have to do it sooner or later ...
>
> Right around the corner is ATI and OpenCL ... Larrabee (?) ... what next ...
>From what the guys were saying on GPUgrid, I gather they were looking into use
>of more than one GPU at a time (for a single work unit), so thats bound
>to break things again.
Agree we need to have a rethink on the whole thing.
MarkJ
<snip>
Enjoy a better web experience. Upgrade to the new Internet Explorer 8
optimised for Yahoo!7. Get it now.
_______________________________________________
boinc_dev mailing list
[email protected]
http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.
