[OpenAFS] the future

Chas Williams (CONTRACTOR) chas@cmf.nrl.navy.mil
Wed, 03 Oct 2012 20:52:28 -0400

In message <20121003203124.GC5179@besserwisser.org>,=?utf-8?B?TcOlbnM=?= Nilsson writes:
>Subject: RE: [OpenAFS] the future Date: Mon, Oct 01, 2012 at 07:28:43PM +00=
>00 Quoting Dyer, Rodney (rmdyer@uncc.edu):
>> Maybe I'm just too old, and in a world where 10 Gb networking is everywhere locality no longer matters.
>RTT to offsite file servers has not improved with fatter links. Caching
>helps a lot. For some things.

that really is the problem.  fatter links only make sense if you can
put enough data on the wire to overcome the bandwidth-delay product.
NFS suffers in this respect just like AFS.

also, if your filesystem protocol requires a lot of individual
transactions to do something like enumerate a directory, round trip
times really start to add up.  so even without caching and just bulkstat,
AFS can wind up being a winner in some cases.