[OpenAFS] cache performance

[Phil.Moore@morganstanley.com]

>>>>> "Nathan" == Neulinger, Nathan <nneul@umr.edu> writes:

Nathan> One assumption I see you making is that your cache isn't
Nathan> flushing.  Unless your cache is larger than your working set,
Nathan> you'll need to run your scans repeatedly to insure that you
Nathan> have a good chance of catching all the accesses.

Good point -- this is another reason we know the cache audits only
provide part of the picture.  I can trivially wipe my cache out just
by running a find | xargs grep against, say, oh how about the openafs
source tree? ;-)

Its also a another reason why the cache audit approach is really the
wrong design, long term.  Again, I'll point out that we did this
because, well, we could.  Had there been a mechanism for getting this
data from the servers, that would be my preference.

Tactically, I'm looking to get better cache audit data, but
strategically, I want to get rid of this mechanism entirely, and do
server-only audits.

In fact, one of the reasons we're looking to abandon the cache audits
long term is that the Windows client is likely going to be installed
here on a frighteningly large scale, and I have no desire to depend on
anything running reliably on Windows... (but I'm a proud
anti-Microsoft bigot :-).