I am trying to connect a new system and it is unable to get any work units. The response says that work is available but my machine cannot finish it within the deadline. While this is a 200Mhz system, which is rather slow, it is quite capable of completing work units within the two weeks allowed.

I don't know if this is related, but although my preferences call for caching from .5 to 1.5 days worth of work, this system is requesting only aout 1800 seconds worth of work (about half an hour). This number seems to be increasing with each attempt (it started asking for about 1500 seconds) so I don't know if it is something odd about the interpretation of the caching parameters or if this system just does not know how many seconds are in a day.

Looking at the benchmarks of that machine, that may be what BOINC is tripping over.

You could try to re-benchmark it.
And set the amount of work cache to between .1 and .5 days. See what that does.
----------------------
Jordâ„¢

Set your caching preference to at last 3 days of work. On my slowest computer, it takes a full day to process a SETI@home WU with a 500 MHz AMD-K6 II processor.

> Set your caching preference to at last 3 days of work. On my slowest computer,
> it takes a full day to process a SETI@home WU with a 500 MHz AMD-K6 II
> processor.
>
It is not nessecary to set the queue size up for slow computers. I have a p166 crunching with a .1 to .2 day cache. There are two possible problems. The benchmarks may be off, rebenchmark. The computer may have spent some time off, that is factored in as well as the benchmarks when determining whether the computer can make the deadline, leave BOINC running for a while, and you may start getting work.

> There are two possible problems.
> The benchmarks may be off, rebenchmark.

Th benchmarks seem reasonable, and have been run several times.

> The computer may have spent some time
> off, that is factored in as well as the benchmarks when determining whether
> the computer can make the deadline, leave BOINC running for a while, and you
> may start getting work.

The system hasn't actually been off, but BOINC was turned off for a while (happened in beta, when the debugging code meant that the then P90 could not meet the deadlines). That sounds like the problem -- the time with BOINC turned off is treated as time that the system is off. That probably also accounts for the small amounts of work being requested -- the factoring probably results in a smaller effective value for seconds per day. That also explains why that amount is going up, as the system is aging out the time off.

Is there any way to reset that statistic? I tried resetting the project, but that didn't do it. I suppose I can try going through some of the XML files and seeing if I can spot anything that looks editable.

> > Set your caching preference to at last 3 days of work. On my slowest
> computer,
> > it takes a full day to process a SETI@home WU with a 500 MHz AMD-K6 II
> > processor.
> >
> It is not nessecary to set the queue size up for slow computers. I have a
> p166 crunching with a .1 to .2 day cache. There are two possible problems.
> The benchmarks may be off, rebenchmark. The computer may have spent some time
> off, that is factored in as well as the benchmarks when determining whether
> the computer can make the deadline, leave BOINC running for a while, and you
> may start getting work.

This is almost certainly why you are getting no work:
>The computer may have spent some time
> off, that is factored in as well

I say that because you mentioned that the amount of work requested was very low and increasing with every attempt. If you leave the computer on with BOINC running normally eventually it should start getting work again. You can also try running predictor on that computer since some of their work is quicker. A clean install will also reset the active fraction so that you can get work. If you want to edit xml look for time_stats - active_frac in client_state.xml and change it to something that resonably represents the percent of the time that computer is running BOINC 50% equals 0.5.

John Keck
BOINCing since 2002/12/08

Yep, that seems to be it. I edited the active fraction, and now it is requesting a more reasonable amount of work. I still need to catch the server at a time when it has some work .....