> One thing that NFS does very well is recover from server failures.
All
> other file sharing protocols that I have seen tend to cause all the
> clients to fail when the server fails.  This feature is the main
reason
> that NFS prevailed over other methods.  Recovering from a server
failure
> was important in the early 80s when NFS was invented because servers
would
> crash a couple of times a week.

Week? Day. 8-)

NFS v2 handled that well (that's what stateless datagram-oriented
protocols are good at), but the introduction of NFS over TCP in v3 broke
some of that resilience. When you lost the state of the TCP connection,
it created a whole new set of interesting failure modes that had to get
resolved by new NFS RPCs. 

That's one of the things I really don't like about NFS v4. It depends on
a reliable network, much more than before, and there's a lot more state
to invalidate if something does go wrong. NFS v4 over IPv6 is a real
horror -- you'd think you were setting up SNA sessions or something
given all the layers. 

----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390

Reply via email to