[OpenAFS-devel] Patch to implement default tuning proposal discussed a while ago

[Steve Simmons]

On Sun, Aug 14, 2005 at 11:45:15AM -0700, Russ Allbery wrote:

> Jim Rees <rees@umich.edu> writes:

> > Those of you with giant client caches, I'm curious why you are doing
> > this.  My own working set seems to be around 50-100 MB, and that's with
> > almost everything in afs (/usr/X11, /usr/local, ...).

> When doing some forms of computation and statistical processing, it's nice
> to be able to load a GB of data set into the local cache and keep it
> there.  One of the major uses of AFS at Stanford is for large data sets
> and running commercial analysis software on it.

Same issue here, Jim (Umich School of Public Health).  With small
caches, one single datafile can cause the entire cache to turn over
with fairly significant effects - especially on shared compute servers.
-- 
Sturgeons Law:      "90% of everything is crap."
Simmons' Corollary: "10% of everything is worth looking at."