Re: desmonitor 0.8 available for beta test

Justin Dolske (dolske@cis.ohio-state.edu)
Sat, 19 Apr 1997 13:23:47 -0400 (EDT)


On Sat, 19 Apr 1997, Howard Cheng wrote:

> Would it be better if you use suspend instead of kill? This way, you
> wouldn't lose the work you have done on the last block. This can be
> a problem if the frequency between busy and idle is relatively high.

Absolutely. :-) I've also got a set of Perl scrips that I've been using to
run stuff at OSU. I know a few people have been asking for them, and I
plan on putting them out on my webpage asap, I've just been so busy...
I watch load a but differently, though -- I jsut look at the output from
"uptime". Same effect.

> On the other hand, suspending the client may increase the time it takes
> for the client to respond to the server, so that the key block may have
> timed out.

I track the number of back-to-back "sleepy" periods. If they outnumber
recent "run" periods by more than a certain threshold, the client is just
killed.

Justin Dolske <URL:http://www.cis.ohio-state.edu/~dolske/>
(dolske@cis.ohio-state.edu)
Graduate Fellow / Research Associate at The Ohio State University, CIS Dept.
-=-=-=-=-=-=-=-=-=-=-=-=-=- Random Sig-o-Matic (tm) -=-=-=-=-=-=-=-=-=-=-=-=-
And then, the world blew up. (Far Side cartoon)