Coda File System

Re: Weak connectivity between servers

From: Robert Watson <robert_at_cyrus.watson.org>
Date: Fri, 3 Apr 1998 13:43:19 -0500 (EST)
On Fri, 3 Apr 1998, David C. Steere wrote:

> Satya'll probably crucify me for this...
> 
> Another way to solve the same problem is to use the ficus approach: the
> things you call servers in your diagram are really clients (use the coda
> client protocols to keep consistent), and export their files to other
> clients. Ficus used NFS to export stuff from their client/server to the
> real clients, we could do that or hack the coda servers to export files
> from their filesystems and not from the special coda partion as is
> currently the case.  (The other drawback of their approach is that they do
> not distinguish between clients and servers, which makes it hard to do
> administration).
> 
> I posted this question a while back, has anyone had time to consider it?
> (Having Coda servers export local files, nfs files, cdrom files, etc).

There has been discussion of exporting read-only file systems (such as
CDROMs) since it would be easy to dynamically create version vectors
(there are few if any changes :).  One idea I have been interested in is a
tertiary storage manager -- i.e., a large repository of mechanically
accessible tapes, etc, and Vice keeping a cache of those files on its hard
disk.  When there is a cache miss, it pulls the file out of tertiary
storage and makes it available to the Venus requesting it, either
transparently, or depending on latency with a 2-step process where the
request is acknowledged but postponed, and then later provided.

Dealing with relication here would be an interesting challenge.

My concern with NFS-exporting Coda file systems is the largely different
semantics of authentication, authorization, not to mention vice cache
misses or running into a conflict that requires user intervention for
resolution.  I suppose with an entirely user-land nfsd, this would be
possible.  I know that this is used with AFS with varying degrees of
success.  In the long run, however, it would be nice to see support in
Coda for doing this directly -- I suspect that the scenarios described in
my previous email are far from uncommon now, and with the increase in use
of mobile computing in various forms, they will become no less common.

  Robert N Watson 


----
Carnegie Mellon University  http://www.cmu.edu/
Trusted Information Systems http://www.tis.com/
SafePort Network Services   http://www.safeport.com/
robert@fledge.watson.org    http://www.watson.org/~robert/
Received on 1998-04-03 13:46:34