Hi all,
I just realize that it's possible that Berkeley has trouble for calculate credit;
but i Think IMHO (as you say ;)) that the lack of credit makes everybody grow their cache i've read on some post that one get 40 WU; another say he his 15 days cache full. don't forget that if you have a 15 days cache; at a T-Time you get WU which the last will be calculate 15 days after (and of course if you don't turn your computer off!). So if everybody has great cache, it's logical that credit makes so long to be calculate.
Moreover, with great cache and the fact that predictor has systematically lower completion date I realize that seti WU were calculate always in the second plan (all predictor, one seti update and once more all predictor...)

So I lower my cache and detach predictor as far as I have seti WU. I think everybody may think about this and determine what they have to do.

Thanks for reading and excuse my poor english (hope you understand what I mean)

Later
Jmdish

> but i Think IMHO (as you say ;)) that the lack of credit makes everybody grow
> their cache i've read on some post that one get 40 WU; another say he his 15
> days cache full. don't forget that if you have a 15 days cache; at a T-Time
> you get WU which the last will be calculate 15 days after (and of course if
> you don't turn your computer off!). So if everybody has great cache, it's
> logical that credit makes so long to be calculate.

Exactly. Not to mention that a giant cache could cause some of the WUs to expire because you have too much work.

> Moreover, with great cache and the fact that predictor has systematically
> lower completion date I realize that seti WU were calculate always in the
> second plan (all predictor, one seti update and once more all predictor...)

It all depends on the expiration date. Predictors are widely variable. I've had some "batches" that jump all over the queue. A couple Predictors, a S@H, another Predictor, etc...

> So I lower my cache and detach predictor as far as I have seti WU. I think
> everybody may think about this and determine what they have to do.

If you feel you must detach from a project, disable network access to process all the WUs you have, then re-enable network access to report your work before detaching. This way you won't toss out WUs that people are waiting on.

<a> [/url]