FW: [OpenAFS] AFS performace question.

Peter Popovics pop@dtv.dk
Sun, 18 Feb 2001 19:42:22 +0100

> >    (I have done some tuning here: I have decreased the
> >    chunk size to 4k, statcache to 1024 entries, and the 
> my experience with unix is that increasing the chunksize 
> actually yields
> better performance.  the default chunksize is 64k, i find that 4m
> performs about 25% better (for files that fit in the cache)

Yees, possibly true in normal case... but there is something wrong... I feel
that the question is not about 25% here...

If I have tried the 32-64k settings with NT (it's more simple to change
these setting there dynamicaly), the speed went down to something similar
than I had with the Linux workstation, less then 100k/sec... However,
reducing the chunk size changed only the write performance, not the
reading... It's not too clear for me...