[2111] in Info-AFS_Redistribution
Re: ptserver stats
daemon@ATHENA.MIT.EDU (Daniel Edward Lovinger)
Wed Oct 20 16:12:08 1993
Date: Wed, 20 Oct 1993 14:50:56 -0400 (EDT)
From: Daniel Edward Lovinger <del+@cmu.edu>
To: Info-AFS@transarc.com
In-Reply-To: <9310201709.AA10526@fish-license.ifs.umich.edu>
Bill Doster <billdo@umich.edu> writes:
> What tools are available to determine ptserver performance,
> workload characteristics, etc? Does CMU have some special
> tools that they use? Other sites?
Other than afsmon, nothing. For a little while I had some logging of
the getCPS calls but the sheer quantity of data we were getting made
it unuseable, on the order of 50-60meg per server per week. Processing
time alone was approaching half a day, and archival was out of the
question. Generating the data was also logically congesting I/O
bandwidth at the same time we needed the fastest I/O (and getCPS is
hit *very* frequently in a large cell).
[ getCPS is the call a fileserver makes to the ptserver to check group
memberships and is thus the logical place to note cell accesses ]
dan
cmu comp services
del+@cmu.edu
