Gregory Mirsky writes:
Perry, may I ask why you exclude Bay and 3Com? Any particular bad
experience from the previous life?
I've only actually had bad experience with Bay stuff -- 3Com isn't
even in the running for the most part -- the software doesn't even
pretend to be able to do the job for a backbone.
And Rob, it you're building a backbone I'd add Ascend's GRF to the list.
Actually, I rather like the architecture of the GRF, though I must
admit to have never tried one out. I've heard some good things and
they might be worth a test or two.
Perry
Well, the GRF has its good and bad points. I've tested one rather
extensively, although I admit it was some time (~8-9 months) ago.
I've been rather upset with Ascend over their lack of reaction to the bug
in the Pipe 150 that had it publishing ARP statments for every ip address
that went by its ethernet interface. Have you found their other products
to be better supported and safer to fire and forget?
jlw
Well, I got rather, uh, pissed at the MAX 4000s desire to publish both a
/32 and a /29 route for all OSPF announcements on dial interfaces (which
went unaddressed in the code for literally months) - particularly troublesome
when you consider the limited RAM in those boxes (and the consequence of
running out of it - it would just drop the OSPF process entirely!), not
to mention a direct violation of the OSPF specifications and the cause of
many complaints from other equipment which this generated.
I've heard they have cleaned up their software act in the last several
months; other than P130s as customer routers for DS1 users (of which we have
a boatload deployed) I have zero *current* operational experience with their
equipment, so my knowledge base on them is ~6-9 months old.
Then again, I'm a SOB when it comes to standards complience, especially when
lack thereof breaks something that we *NEED* around here (such as reliable
service :-).
I still don't like CISCO's RAS implementations, but I have to say this -
for all their warts, including some business policies that I consider
nothing short of INSANE, their router hardware and IOS still win the prize
for uptime in my experience.
A real example from our core:
XXXXXXX-CoreX uptime is 38 weeks, 1 day, 7 hours, 49 minutes
System restarted by power-on
That's pretty typical around here; the last "power on" was to do routine
maintenance on that particular device. 
An anecdote that is related to this...
In January, on the day we (NACS.NET) had our open house, the RSP-2
board in our 7513 (which we bought used) died.
Cisco worked with us to get a replacement in four hours, even though they
were only contractually obligated to 24 hours.
Kudos to Cisco's Cleveland office for going WAY beyond the call of duty.
And when I'm ready to buy a dedicated router, it WILL be a Cisco.
(Oh yeah. Other than that, no downtime issues yet.)
Jason,
About a year ago I heard a rumor that a pretty good sized batch of
Ascend P50s made it out of the factory with the same MAC address. Of
course, this is semi-OK if they all go to seperate sites, but the Max 1800
was not in widespread use at that time, so the only option for doing
dedicated ISDN BRI <-> BRI was to have a rack full of P50s. You can
imagine what fun the poor guy that had to figure that one out went
through before he figured out the real problem, called Ascend, and
explained to them how ARP works...
Blake Willis
CAIS Engineering
i like the GRF 400 i have running in NYC pretty much, except for the
terrible documentation. ascend tech support tells me a new, improved manual
is on the way -- it'd definitely needed. some of the stuff in the manual
they shipped me is simply wrong, and i've needed to pry correct information
out of their tech support organization a number of points.
i also agree with the other comments about ascend's software release group;
they really need to get their act together.
on the other hand, the port density and the cost per port on the GRF are
both quite good, and the performance is definitely there. we're running the
fast ethernet card, OC3c card, and FDDI card in ours, and they all are
gettingg the job done.
richard