2010-02-18

Truth

But after HP integrates 3Com, they have the whole portfolio too. In fact, not a very good one since 3Com equipment is not exactly recognised as quality. Most importantly, the people who design networks and lay out strategy remember being completely shafted by 3Com in the early 2000’s and they haven’t forgotten.
( Source )

It's true. In the late '90s when we deployed CoreBuilder 3500, SuperStack 9300 and 3900 switches as edge devices, they worked as advertised. We still have some 3900 switches in service and really the only two reasons they are being retired is 1) they don't do POE and 2) in our experience 3Com legacy hardware support totally blows[1].

We had some disquiet when we saw the 4300 family be released, they appeared to be a step backwards in a lot of ways. The 4007 switch was a big heavy boat anchor. The 4005 was a pretty good switch and probably did well as routers in general, however we had one in a hardware developer's network and it had some interesting problems[2]. When we tried to scale up, the 4005's limitations started to become restrictive, and now we have almost no 3Com gear in our network save some basic 3900s.

Since then our experience with 3Com has been less than ideal -- one customer tried to use a 4500G switch as a core router, but its limits prevented that and it had a really weird way of doing ACLs. The step up, the 5500G, was weird in its own ways.

So yes, we remember. Now we buy Cisco for core services, with Linksys or Dell for the edge. And because Dells have 3-year warranties, and are cheap enough to buy two of, we just use cold-standby as our switch-failure coverage.

--

[1] : several years back, we had 24/7, four-hour onsite advance replacement coverage on a switch that was pretty important. It cost an unreal amount of money because the switch was getting on in years. So the switch dies, and the replacement switch fails to show up as per our contract. It turns out that 3Com had sold the support contract to a company which basically did nothing except collect our money. Eventually someone had to ship us a switch via air from Toronto. Four hour onsite? Ha, ha. That 3Com switch is gone now.

[2] : Their product did IP-failover. Problem is, the 4005 didn't update its arp cache if contradictory information was seen on the network, and it took ages for the cache entries to expire on their own. (I seem to also remember that you could set arp cache timeout manually, but the switch ignored those settings.) So when the failover happened was simulated, the IP was reachable in the lab but not across the 4005s' router. First time a bug we reported to a vendor triggered an entry in a firmware update release. Oh, and the telnet interface to the management brain kept dying too, so you could only talk to the switch through a serial connection.