From: Niall O Broin (niall at domain linux.ie)
Date: Mon 13 Aug 2001 - 16:56:42 IST
On Mon, Aug 13, 2001 at 02:14:01PM +0000, Conor Daly wrote:
> Naturally. Of course, this is getting away from the main point. The
> choice of using hub/switches in each computer lab and a single cable back
> to the server or individual cables for each client really depends on the
> x-terminals' resolution, bit depth and frame rate. If these are low, a
> single 100Mbit link per lab would probably carry all the traffic but if they
> start to climb, then individual links are required.
X terminals do hit a network hard, but many of us have run multiple X
terminals over a single 10Mbit network and lived with it. As you are going
to have 1 100Mbit link to your server, then it really doesn't matter what
way you have the clients, given the number of clients you're talking about
and the layout you describe. If you put switches in each lab and a link from
each of them to the server, then you force congestion to happen at the lab
switches. If you have one switch, then congestion happens there. In any
event, unless you go for a switch with a turbo backplane and a Gigabit
channel and Gigabit connection to the server, which is probably outside the
budget, you've got 100 Mbit to the server and that's your bottleneck. Quite
honestly, I wouldn't consider it much of a bottleneck in the environment
you're describing.
I'd definitely go with one switch for which I'd have a spare. It'll probably
never fail, but having the spare on the shelf beside it more or less
guarantees that it won't fail :-) Mightn't be a bad idea to have a spare
server too. As that's probably outside the budget, see what you would need
to do to promote one of the clients to a server, because of course in this
environment if your server goes down, you've lost everything, and I imagine
on-site service will be slow.
Niall
This archive was generated by hypermail 2.1.6 : Thu 06 Feb 2003 - 13:11:35 GMT