At 10:07 AM 8/26/99 -0400, Morris Strongson wrote:
> Thanks for your replies. Russ' answer has me even more curious
> as to whether a copy of the whole file sits in the cache? I
> would tend to assume not from his answer, but I'm hoping to get
> a definitive answer on this. We have users who will be
> loading libraries up to one gig.
AFS (and DFS) do read and write operations on byte ranges, by
default in 64k chunks. Typically the Cache Mgr will retrieve
those ranges requested by an application, so if an application
requests file portions delineated by byte ranges then that is
what will be retrieved. In the case where an application
retrieves a file that is larger than a cache I believe what will
happen is the file will be read into cache in chunks,
correspondingly fed to the application in chunks, and then as
the cache is filled up those chunks already fed to the
application will be replaced. Of course if you will often be
using files larger than your client cache size you may want
to start tweaking client and server parameters for performance
tuning, depending on your network (e.g. chunk size, afsd
arguments, etc.).
=======================
Tom Menner
Technical Consultant
IBM Transarc Laboratory
work (412) 338-4414
page (800) 386-6908
