On Friday 27 August 2004 17:35, Sam Varshavchik wrote:
Jerry Amundson writes:
On Thursday 26 August 2004 05:44 pm, Sam Varshavchik wrote:
Or, it's possible that a fractional T1 between the two offices may
be cheaper than a pair of backup low-bandwidth circuits. Except
that a backhoe incident in the immediate neighborhood will likely
take out both circuits.
I should clarify - it's been mandated that I put a server in the
remote location... or else we move to Exchange.
Folks in the remote office, about 25 users all using IMAP via
Outlook 2000, were seeing intermitent response degradation, the
cause of which I have been unable to pinpoint. The frustration
level has gotten pretty high there, so it has been decided they get
their own server.
Oh, a fractional T1 between the two offices will be immune to the
general hiccups on the Internet. You'll have to get the tricky
routing working, but that's going to be the worst of it.
Both ends have full T1's, with IPsec doing the VPN between us. I've
strace'd, tcpdump'd, etc. but it's never been clear what causes the
hesitation, and of course it never happens consistently. I've always
suspected some kind of Outlook-IMAP_ENHANCEDIDLE-famd thing, but have
been unable to prove anything.
You can still put a server there, and have both servers cross-mount
each other over NFS.
But, essentially, the maildirs of all users need to physically exist on
both servers to attain redundancy. The Coda suggestion seems worth
looking into, though it may make things more complicated than
I had also been thinking NFS early on, but over a T1? If I determine T1
bandwidth is adequate, it may simplify things.