Martin,

On Thu, 05 Aug 2004 02:18:44 +0000, Martin 53N 1W wrote:

Randall Schulz wrote:
[...]

1) The SETI@home client tends to "stall" occasionally--maybe once each
day--and must be manually stopped and restarted to get it going again.

Never suffered that, on either Windows or Linux machines.

2) The extrapolation of the remaining time is non-linear, indicating
that progress near the beginning and end of the work unit proceed much
faster

[...]

Yes, this is well known and can be calibrated out in such as SetiSpy for
monitoring s@h classic.

Yes, I guessed as much. I've noticed the pattern seems the same on both
Windows and Linux.

Has anybody else noticed these phenomema? Does anyone know why they
happen? In the case of (1), is there a way to prevent this problem?

Machine or CPU overheating? Overclocked? Choked with dust? PSU on its way
out?? Other bad hardware or a bad 'interaction' with whatever Windoze
virus scanner??

Complete sentences?

As I said, I'm running Linux.

This symptom has occurred on two different versions of Linux (two
different distributions and two different kernels) and on two different
hardware setups (motherboard and CPU). The power supply has ample excess
capacity and is fairly new (about a year). The previous hardware was
mildly overclocked and displayed no other problematic symptoms, and the
new one is not overclocked at all. (The old hardware died because I
neglected to clean the dust out of its heat sink--I couldn't believe how
much was in there! I'll never let that appen again...)

The other thing I neglected to mention is that KSetiwatch always showed
70% completion at the point the WU stalled. That was on the old CPU. On
the new one, where I've only experienced this once so far, it was 68%.

Weird, eh?


It's all pretty much moot, anyway, since I'm switching to BOINC. I'm only
running classic as a stop-gap when BOINC can't cough up any work units.

Good luck,
Martin

Thanks. So far it looks like more luck is needed in the care and feeding
of BOINC than Classic.

Randall Schulz