[18974] in Athena Bugs
Re: Netscape 'save to disk' doesn't get along with AFS.
daemon@ATHENA.MIT.EDU (Greg Hudson)
Thu May 24 11:47:59 2001
Message-Id: <200105241547.LAA08847@egyptian-gods.MIT.EDU>
To: Bill Cattey <wdc@MIT.EDU>
cc: bugs@MIT.EDU
In-Reply-To: Your message of "Tue, 22 May 2001 21:12:01 -0000."
<ov2hMVRz00018L4oRT@mit.edu>
Date: Thu, 24 May 2001 11:47:56 -0400
From: Greg Hudson <ghudson@MIT.EDU>
> Everyone knows that stat operations must be slow and time consuming
> on a distributed filesystem.
Oh, another reason you're losing (besides the afsd parameters, which
we discussed out of band) is because in 8.4, our AFS clients don't
properly use the "bulk stat" optimization, so instead of reading all
222 directory entries in one operation, your machine has to make a
round trip to the AFS server for each file.
In 9.0, we're using AFS modules we built ourselves with the bulk stat
bug fixed. So performance for this case should be much better.
(Statement applies to all platforms.)
