Does the afsd "chuncksize" parameter determine the actual size, or the
maximum size of data chunks handled by the cache manager? That is, can
a data fetch (or store) transfer fewer than "chunksize" bytes, or is the
size of the transfer buffer fixed at "chunksize" bytes?
Stated in terms of an example, suppose an AFS client requires frequent
access to a large pool of 1K files. Under these conditions, would
reducing the default (64k) chuncksize affect cache performance, or does
a 1K fetch transer 1k bytes, regardless of chunksize? If so, is there
ever a reason to reduce chucksize? Any reason not to make it bigger?
Thanks for any insight,
Bob Dew
National Institutes of Health
Bethesda, MD
