[tech] Re: UCC uplink network upgrade
Nick Bannon
nick at ucc.gu.uwa.edu.au
Thu Jun 19 15:34:02 WST 2003
On Thu, Jun 19, 2003 at 12:36:37PM +0800, David Cake wrote:
> At this stage, I am going to try for 4 core fibre, with the
> second pairs purely as a backup.
That sounds like a good decision.
I talked to GR Services and no, blown fibre isn't going to work. The
cable is going to have to go through a 20mm conduit and the 2 duct
tubing (plus cladding) is 16mm, which is too close. They're going to
have to pull out the 10Base2 RG-58 while they're at it.
Apparently the maintenance guys are unhappy about the conduit as well,
but we're kinda stuck with it because there's others wires running
through it, too.
So... we get one cable and if we ever want to upgrade it'll have to get
pulled out and replaced, too.
It's tempting to go for high quality singlemode in that case and have
overengineered 10km range 1000Base-LX links, but compatible with the
other media converters that UCS is putting in. The installation cost
will be similar (less for the cable, more for termination), but the
endpoints would be 3x$797 dual 1000Base-LX modules instead of 3x$493
dual 1000Base-SX modules.
> I am undecided about Alloy vs DLink
> switches - I wish I had more idea about how the various switches will
> actually perform in action. The Alloys in particular are a bit of an
> unknown quantity.
> Reliability is my main concern about the switches, providing
> we have the requisite VLAN support etc.
Right.
> I don't care much about 100
> vs gig ether (though it seems important to Nick, I think there is a
> very little to gain from it), or about minor cost differences.
> Cheers
> David
I think that we've "needed" 100Mbps for a couple of years now. One can
always make do, but lots of things, almost entirely outside of the
UCC's control, have to line up to make any kind of upgrade possible.
(e.g. simultaneous availability of upgrade funds for UCS, Guild and
UCC, willingness to allow a major network outage, and a shared opinion
that it's "worth it".)
So, given that this upgrade's going to have to last a while it seems
worth paying a bit extra to get ahead of the curve. When it turns out
that backbone gigabit is now so mainstream that it's a 8% _saving_ over
the original DLink quote then it seems like we'd be mad not to. Even
sticking to DLink it's still "only" a 50% extra cost per switch,
because they're following the Cisco model of stiffing you on the
optional modules and media converters. Upgrading later down the track
means paying for all that, convincing everyone it's worthwhile again,
_plus_ having no guarantee of finding the exact correct module for a
two? year old switch by that point. More likely, it means replacing the
switches. What's a fibre module's MTBF, anyway?
I'm personally willing to pay the upgrade costs from the original
"adequate" solution to the "luxurious" 1000Base-SX/LX solution, though
I do hope that others help out. The UCC can't afford to stick a lab's
worth of $2000 desktops around the clubroom every year or two, but it
can afford cool shared servers and a cool shared network.
Nick.
--
Nick Bannon | "I made this letter longer than usual because
nick-sig at rcpt.to | I lack the time to make it shorter." - Pascal
More information about the tech
mailing list