Odysseus <odysseus1479-at@yahoo-dot.ca> wrote in news:odysseus1479-at-
A55B81.05032019072009@news.telus.net:

In article <iLx8m.27974$ay4.15498@newsfe27.ams2>,
 Skywise <into@oblivion.nothing.com> wrote:

<snip>

I've got my "connect about every" and "additional work buffer"
both maxed out to 10 days on each machine. The hope is, I can
keep a LOT of work in the queue. I never do get a full 20 days
work.

It's not recommended that you set your "connect every" high when your 
system has continuous (or frequent) access. That was the best kluge 
available in the days before BOINC supported customized caches _per se_, 
but now that the "work buffer" setting is available, the scheduler takes 
the "connect every" more seriously. In particular it will refuse work 
with deadlines that come sooner than this interval, on the assumption 
that the host won't be back online again in time to report the results.

So what is suggested? Connect every 1 and additional work 10?

For quite a while I was stuck on dialup and managing to keep the
quad/cuda system fed required leaving the system connected all
night to download enough tasks. At times, the tasks were burning
up as fast as I could download them. It's better now that I have
some high speed.

But we still have the issue of getting new tasks if there are
pending uploads. If the upload servers are borked, we can't
clear them out to be allowed new downloads.

I don't have bleeding edge hardware, but ever since I added my
quad/cuda system, I've been climbing the charts quite nicely,
until all the woes the past month or two.

Brian
-- 
http://www.skywise711.com - Lasers, Seismology, Astronomy, Skepticism
Seismic FAQ: http://www.skywise711.com/SeismicFAQ/SeismicFAQ.html
Quake "predictions": http://www.skywise711.com/quakes/EQDB/index.html
Sed quis custodiet ipsos Custodes?