> We could add two relatively small and inexpensive non-RAID servers which
> would both have Firebird installed in identical configurations. One of the
> machines would not have the Firebird service running most of the time. Both
> servers would be configured to use a third (larger expensive RAID) machine
> as their data drive. I figure putting two gigabit ethernet adapters in the
> storage server, and gigbit ethernet in the two FB machines would give
> reasonable network performance.

I do not know what data you are dealing with, but given the
volumn, I would be looking at just how much is changing. It
is only changes that have to be replicated, so an ARRAY of
computers all running Linux and Firebird whould also help
load share. With the bulk data exising on at least two
machines. Add a 'last update' timestamp to records, and the
management end need not be too complicated.

> Theoretically, if the primary FB machine goes down (or requires
> maintenance), the FB process could be stopped on that machine and started on
> the backup FB machine for a rudimentary fail-over system. Granted, the
> storage server can still be considered a "single point of failure", but my
> network admin figures a file server running no major application process of
> its own is easier to keep running 24x7 than other machines.

There does not need to be a primary machine - but again it
depends on how the data is being updated. I have a network
that can't be relied on ( British Telecom supply it <g> ) so
ALL machine run Firebird, and tick away locally while the
network is out, probably not practical in your case, but
it's just an example of 'lateral thinking'. Does everybody
need access to everthing all the time? Can some tables be
read-only most of the time?

--
Lester Caine
-----------------------------
L.S.Caine Electronic Services