^^^^^^^^^^^^^^^^What he said^^^^^^^^^^^^^^^^^^^^^^^^

Exactly what am I doing here? I only have one option to control connection frequency, and it's not defined in hours per day:

• Connect to network about every ____ days
  

How do I use that to control it according to what you described? I've actually had that set for 0.025 (days), because BOINC was in the habit of accumulating piles of completed work units, which caused ridiculous swings in my RAC and made it impossible to get a sense of general effectiveness. I am loathe to undo this and wind up having useless averages again.

As far as process priority, I already have a licensed version of Cacheman XP, which also has "sticky" priority settings.

Actually I use the method described in this FAQ for the scheduled connecting i.e. doing this through local preferences rather than via the website device profiles as that allows daily variation:

Some sample Scenarios when to Crunch or Network
http://www.worldcommunitygrid.org/forums/wcg/viewthread_thread,20741

RAC/BSRAC is a sliding mean of the last 60 days or rather the average of the days that results were reported in that 60 day window, so it should not flux that much, but the logic is flawed if there is a big outage day and result reporting slips to the next day (and BOINCStats is not in sync with the WCG stats time keeping). At constant production it should not matter that much and at least I see it move only so slightly, more dependent on what gets OKayed on from prior pending validation status. On Sunday I had 34 in PV, and Tuesday morning it was 19... and that's with 1.5 devices.

Also if using a scheduled connect of an hour, most/all Ready to Report will be cleared during that 1 hour daily window. BOINC forces any out that have reached 24 hour after completion and when it does, it takes the newer along, since there will be several requests for work and these combine with Ready to Report clearing.

It seems like Sekerob and XS_fallwind are using two different means to achieve the same thing, the latter being the more geekish method I suppose. I've been using a cache of 3 days since my SETI@Home days; it was motivated by the legendary outages that project used to have, and I confess I have yet to see a similar outage with WCG. I still use the large cache anyway.

@Sekerob: Why enable the "Disconnect when done" setting, since clearly you're using broadband? I didn't think the setting was relevant in a true networked context, and I worried that it might even force an unnecessary renewal of the DHCP lease.

I'm an ignorant American who hasn't yet got up to speed with UTC. What values represent noon and midnight in UTC? I've known about boinccmd but never felt inspired to use it before.