10 Gig is enterprise-ready, but Cisco doesn’t play well with others

Click on over to InfoWorld’s 10-Gig switch-off special report, and you can read about our latest adventures testing 10-gig switching solutions at the Advanced Network Computing Lab, now nestled comfortably in its new digs at the University of Hawaii.

When we started planning for this test, we invited eight separate vendors, including 3Com, Enterasys, Extreme, Foundry, Force 10, HP, Nortel, and, of course, Cisco. We even got Microsoft and Sun to weigh in on their 10-gig operating system goals in two podcast interviews you can find right here.

I knew Extreme was in because those boys never run from a fight. They don’t have to. HP was a surprise in that I hadn’t thought of ProCurve and enterprise iron fitting in the same sentence, and the company was set to prove me wrong big-time.

Nortel politely ignored the invite as usual. Foundry and Force 10 both considered our test plan, but flat out said it wasn’t enough of a “bit blaster” test. Both these companies have heavy carrier orientations when it comes to selling their 10-gig solutions, so what they want to show in a comparative is loads and loads of traffic blasting through the switch.

That’s fair, but it’s not what enterprises care about, which is InfoWorld‘s orientation. More important to an enterprise CIO is whether 10 Gig is really stable. Loads of man-hours spent tweaking switches, network cards, drivers, and the like only to get just the most out of a 10-Gig pipe simply isn’t good enough. They want to know if the solution works. Do they get 10-Gig performance simply by powering up? After that, they care about easy integration into an existing architecture and day-to-day management tools, and they certainly have budget concerns. But most of all, they want to know what’s in it for them besides just a really fat pipe. What’s 10 Gig really good for that straight Gigabit Ethernet doesn’t already give them?

Our previous 10-Gig test didn’t have many good answers to any of these CIO-style questions. The technology was really a carrier class operation back then. Plus, the killer app hadn’t yet fully emerged. To my mind, that killer app is a combination of VoIP and storage with computing clusters close behind. These are fully mature technologies in the case of the first two (soon for clusters), and 10 Gig is the pipe you want to run them.

SANs have obvious 10-Gig benefits, whereas VoIP finally achieves its promised purpose. Even today, many VoIP networks are still completely separate from their data counterparts because latency and capacity issues simply aren’t yet there under gigabit networks with large call volumes. But this test clearly showed that today’s 10-Gig solutions not only have plenty of pipe room to run your voice traffic over your data pipes, their management tools also have the smarts to make that operation doable for mere mortals rather than staff-toting VoIP Gandalfs.

Finally, what grabbed me about both our contenders was price. Looking at the article, you might think that HP’s ProCurve was the real standout here — and you’d be mostly right. HP’s overall solution price was amazing considering the technology we were testing — a testament to its all-enterprise-all-the-time sales goals. But the company also didn’t really bring all the hardware it needed to scale to the full solution we described in our test scenario. The reps really just brought along what they needed to get through our performance tests. Even if they had brought what they needed to scale, however, the solution still wouldn’t have cost much more than $150,000, and that includes a bunch of wiring closets. That’s an eyebrow-raising price tag by anyone’s standard.

Extreme brought all the big iron and had a price tag to match. The company’s solution, however, could have scaled to the moon and back without breaking a sweat. But even here, the move from carrier to enterprise is visible. Extreme didn’t bring any of its lower-cost Summit switch line, which would have brought its wiring closet costs down quite a bit. The company also has smaller core switches than the BlackDiamond 10808 that it brought, which also would have affected cost. They even have a smaller core-capable box aimed specifically at smaller enterprise customers.

Somewhere in here is most likely where we lost Cisco, which had agreed to participate in our roundup, then inexplicably backed out at the last minute. The company just doesn’t like technology tests focused on areas where it hasn’t quite gotten its marketing message straight. I once had a Cisco exec explain to me on condition of anonymity why the company rarely, if ever, participated in comparative shoot-outs. “We’re number one,” he said. “So the best we can hope for is validation of something that’s already known. Other than that, we can only lose.”

I chewed on that, and decided he had a point. Still, the company complains whenever it’s not invited to one of these, so we send the invites.

I just wish Cisco had the southerly nodules to simply refuse straight-up like Foundry and Force 10, rather than stringing us along, sucking up a test slot that it turned out 3Com could have used, and then dropping out at the last minute when it was too late to get a replacement. That kind of behavior robs me, and it robs you of reading an article with an extra competitor. Still, the rest of the test went well and a day spent testing in paradise is still better than my usual. So check out this week’s 10-Gig excursion and start thinking about a move to a fatter pipe: The time is definitely ripe. Mahalo.