| Subject: Re: My Over-commitment solution (detach from climate prediction) |
| From: "Stratcat" <none@no.org> |
| Date: 25/09/2005, 05:54 |
"tjs" <anon@anon.net> wrote in message news:43296e50_4@news1.prserv.net...
Recently while Seti was having their server problems I was running
Einstein
100%. I decided that this was a good thing so I re-attached my Boinc
manager to
LHC and ClimatePrediction so I could always have work going, unless all 4
were
down! A few days later as my various completed work units were needing
uploaded,
I noticed I was not getting new work units from anyone. Climate prediction
does
not get new WUs since a unit is so big (6 months to complete on my
system). No
mater what I tried I could not get Boinc Manager to ask for new WUs.
I read about BoincDV which can show you internal accounting info for you
specific Boinc manager and the workloads you have banked up, and I guess
the
backlog of work your commited to. BoincDV can reset these values.
Supposedly if
you zero out the deficits You can fix some problems. I tried this
(although
there are warnings in the process) and I got some work units, but all was
not
fixed.
All this started after I attached to CPD. This morning I detached from CPD
and
immediately Boinc began asking for Seti, Einstein and LHC WUs. Apparently
the
nature of Climate Predictions huge single work unit style, causes the
Boinc
Manager to report that you PC is fully commited to that work, even though
it
only gets 25% of the computing time.
Additionally I notice system instability with the Boinc Manager while
running
CPD units. I had the Boinc Manager freeze up as far as the GUI interface
goes
(could not click on it to bring it to front) and had to kill the process
several
times to unfreeze the computer.
Maybe this will help some others out there. I never have read about this
effect
in any FAQ or posting before.
Alternatively, you could have "suspended" CPDN, and forced "updates" of your
other
projects, then "resumed" CPDN after the other projects D/L'd thier work.
BTW, 6 mos is an extremely long (though not unheard of) time to complete a
CPDN
model, unless you mean at the 25% resource allocation (meaning ~6 wk @
100%).
Strat