On Sep 29, 2009, at 11:07 AM, john.mcl...@sybase.com wrote: > It has to wait until the first task is complete to get any estimate. > Neither 95% complete or an estimated 5 minutes remaining will do. > > There is a project currently in existence (Drug Discovery) where the > progress through % complete is extremely non-linear with respect to > time. > The first 96% is done in < 1 minute, and the last 4% takes an hour > or more > (on my machines, your mileage will vary). > > We still have projects that go to 104+% and then spend quite a bit > of time, > possibly bouncing back to around 95% before completion. > > The first task from a project would have to run at high priority as we > would have no idea at all how long it was going to take. >
And why is that a bad idea? On initial fills of the queue it is far less dangerous to actually run tasks to completion on project initiation until it is known how long the tasks will take. It would be far easier to explain under fill than the almost certain overfill most systems see on first attachment. This is a very common complaint where because of the DCF issue as Richard noted the participant gets days to weeks worth of work until DCF starts to settle in. The only place where this rule is inverted is on some projects with GPU versions where you WILL under-fill until DCF settles ... _______________________________________________ boinc_dev mailing list boinc_dev@ssl.berkeley.edu http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev To unsubscribe, visit the above URL and (near bottom of page) enter your email address.
