"Rodney Pont" <pmmspamscop@infohitsystems.ltd.uk> wrote:

On Tue, 07 Jul 2009 20:59:44 -0700, DaveT wrote:

At other times, however, it gets into a "state". When I run a manual update, it says "Not reporting
or requesting tasks" even though there are very few available in my queue, or even if I bump it up
to 10 days and update the preferences. It still says "Not reporting or requesting tasks". Then,
eventually, all of a sudden, it "switches", and starts downloading like crazy, and gets a page and a
half of tasks in a minute or so. Can anyone help me understand this behavior? What is it waiting
for?

Has it got any failed uploads it's waiting to retry? I've noticed mine
doesn't even attempt to fetch more work if it's having trouble
uploading work already completed but as soon as it's managed to upload
everything it does start fetching again.

Yes and no. No, there are currently no failed or pending uploads and I am still getting "Not
reporting or requesting tasks", however, I have also noticed that when there are such tasks, the
system won't fetch, and it begins as soon as they are processed and (possibly) removed from the task
list.

This has led me to wonder if there is some sort of "threshold switch" such that if there are more
than a certain number of tasks (in any state) in the queue, the switch is "off", this number has to
fall below a certain value before Boinc will again start fetching, perhaps for a certain number of
minutes until it again switches off. Another arrangement that would explain this is if the SAH
servers had their own setting that said, in effect, "only send so much work based on the CPU
capacity", sort of a form of rationing. These are but two examples in a range of theories that would
explain the observed behavior. 

Another odd thing that happened was that after not being able to get any work since about (last)
Thurs, 7/2, I did an update on Saturday night, 7/4, and whough, a flood of work came down, and in
about 2 minutes, the queue was full. I'd be happy to channel my downloading into off-peak hours if
that would help the bandwidth situation, but I can't seem to control how the system acquires work
because of the above issue. There appears to be two weekly "windows" during which the SAH system is
responsive. One is from the end of business Mondays until the Tuesday shutdown. The other is when
the system recovers from the Tuesday shutdown, usually Thursday or Friday, until it again "clogs up"
on the weekend. 

There are currently 26 tasks in the Queue: 3 processing and 23 waiting to run. Of these, 12 are cuda
and 14 are "regular". The preferences are set to 2.0 days work. The system is currently "Not
reporting or requesting tasks" when I do an update. The 2.0 days of work preference normally results
in about 75 tasks if and when downloads happen freely. Based on past experience, I would expect it
to start fetching (or trying to) when the queue falls to about 20-23. There have been exceptions,
however. If sufficient tasks have not downloaded before one of the above "windows" closes, then I
will tend to run out of work before the next one.

Another possibility is that the "To completion" times, which are currently quite long, have
something to do with it, but that wouldn't explain why bumping up the # days in the perferences (and
updating the preferences) doesn't have any impact. 

Anyhow, that's about all I know about it. I have two other questions. One is, what does "Connected
to localhost (6.6.31)" at the bottom of the Boinc screen mean? It is not a dotted decimal IP address
or a port number. The other is, is it possible to run Seti on other computers and obtain work from a
central Boinc? I used to do this with SetiGate; I don't know if this is possible with Boinc. (Among
other things, I was thinking about depleting the queue when it gets near one of the windows, to
trigger fetching of work.) 

Thanks for any info.
Yours,
Dave