Coda File System

Re: Two questions: usability on large network and K5

From: Ivan Popov <pin_at_math.chalmers.se>
Date: Fri, 6 Dec 2002 15:29:16 +0100 (MET)
Hello Fernando,

> 	I'm evaluating CODA for usage in a large network and two questions:
>
> a) Is it usable in a 3500 workstation network?

I do not think somebody had tested Coda for similar setups.
Coda creates very modest traffic as long as the files can be
cached (i.e. unless there are massive updates on files used by every
workstation), so it should scale very well.

I cannot tell you how well one server deals with a huge *number* of
clients (in terms of resources needed on server per active client).
Replication would help that, if it becomes a problem.

I would guess that Coda scalability is not limited by the number of client
hosts, but rather by some other parameters that depend on your needs and
usage patterns.

The usability depends enormously on the kind of data you put on coda.
Same data updated in parallel on multiple clients would kill your system
(or rather the users would kill the administrator when they face file
conflicts too often :)

Big files are tedious to open on a moderately slow/fast network (Coda
caches the whole file before open() returns to the calling process).

There is a limit on data amount (20-30G) per server process, too.
It makes administration a bit more complicated for bigger servers.

> b) Does it support Kerberos V authentication?

Sure. Coda uses its own tokens, but it can use Kerberos to acquire them.
I am running Coda with kerberos5 authentication.

Be warned that Coda is still "unsecured", probably harder to exploit than
NFS but still vulnerable.

It would be exciting to see a large Coda installatin in action!

My 0.02 euro :-)
--
Ivan
Received on 2002-12-06 09:30:53