[OpenAFS] cache performance

[Phil.Moore@morganstanley.com]

>>>>> "Derrick" == Derrick J Brashear <shadow@dementia.org> writes:

On Thu, 24 Oct 2002 Phil.Moore@morganstanley.com wrote:
>> enterprise wide reports on who is accessing what volumes.  This has
>> proven very useful.

Derrick> Who at a user level or machine level?

Machine, of course, but that's a good place to start.  I'd love to get
the user's, too.  Does the cache manager keep track of this at all?

>> Now, what you are saying suggests that we can't contoinue to make this
>> assumption about the access times, and thus, we'll have to rethink our
>> cache analysis code.

Derrick> Snoop the wire and record the data you want (at the fileserver end)?

Snooping the wire, enterprise-wide, would be prohibitively expensive,
especially in a very large environment under very heavy load.  We
don't care about realtime analysis, we want batch audits that give us
last 24 hour statistics.

As I mentioned in a seperate note, the "right place" to get all of
this information is from the fileservers, and one of the things on my
virtual list (virtual because its in my head, and not written down
yet), is to fund work to enhance the performance and usage statistics
mechanisms in the AFS server code.