Bill, have a peek at Caldera.  I recall a couple of reviews on a 
commercial product from them targeted at Enterprise management of Linux 
desktops & servers.  Ya, it costs USD$$$, but with 700 desktops to support 
you likely already are looking at or you have something like that to keep 
them all in sync with updates and whatnot?

<rant ON>
Single point of failure is a tough problem to solve.  I'll make an 
assumption that "single point of failure" ties with uptime, high 
availability and fault tolerant computing infrastructure.  I've consulted 
for very few companies that every fully grasp the cost of 99.999% uptime.  
Do you want to do this?  I have done it, it is not cheap.  Please 
prepare management for the ordeal of writing many 100,000 to 500,000 USD 
checks to accomplish the task.

Redundant routers, dual NIC'd desktops, double ethernet wiring, 
geographically seperated utility power feeds, ditto UPS & Gen Sets, ditto 
Telcom connections.  Two of everything?  Hmmm, dude, you're getting the 
budget of your dreams!

Better to plan for failure, document and PRACTICE recovering from 
failures.  Eliminating single points of failure can be accomplished, but 
the expense to the business often exceeds the amount of risk.
</rant OFF>

(I'll bet that you are looking to eliminate the single points of failure 
that are financially practical to solve, yes?)

Sun has a very nice server failure capability.  The simple 
configuration is shared storage is used by two "server" hosts.  One is 
hot, one is a warm spare.  The arp'd ethernet address for the IP address 
all clients connect to for their NFS (or whatever?) is able to be moved to 
the warm spare when the hot server fails.  I think Microsoft's kinda works 
in this space too?

There are few products out there to implement this basic server failover 
for Linux systems.  Including SAN style storage you can share between the 
servers.  Considering the past year's dot com bombs, you'll want to do 
some fresh research in this space.  Even the big players are laying off 
people, I would hate to purchase from a company that's on the brink of 
failure.  A really solid SAN that is accessable by multiple servers can 
solve a lot of the data/business continuity headache.  Something really 
HA/FT, like able to lose a couple of power supplies and a handful of disks 
without kicking out your users.

On the cheap side, CODA sounds promising.  Can it reconnect & 
re-authenticate to the warm-standby server?

Can smb or nfs meet the requirement if upon system failure your users 
understand that they need only log-out & log-in to keep working?  (kinda 
like they would need to do if the primary MS file server died and was 
replaced with a warm-spare?)

 - tom e.

On 2 Oct 2002, Bill Warner wrote:

> Yes this is a great topic for the list and for Linux people in general. 
> 
> We have moved almost all of our old SCO UNIX servers to Linux and are
> looking into the possibility of moving our corporate desktops to Linux
> as well.  Our UNIX team has been using Linux desktops for a long time so
> we maintain some of the needed things.  I have been charged with the
> task of evaluating and putting together a working desktop system that
> will fill all the requirements that our win2k desktops do now.
> 
> Some of the things we need are performance and no single point of
> failure.  I seem to have problems with these things when it comes to
> setting up unified logins/home directories.  I know I can setup smb
> authentication.  I could use NFS homes but those both give a single
> point of failure.
> 
> If anyone knows of a good way to support 700 Linux desktops I would be
> more than open to suggest ins.