White-box switching has the potential to change the market, but in ways that aren’t necessarily obvious. Alan Weckl, a Dell’Oro Group analyst, made some good points about that — and about how the Ethernet switch market is changing with or without white boxes — during this week’s Ethernet Technology Summit.
Weckl gave a keynote on Wednesday, and he also spoke during the Summit’s seminar on open-source switching. I’ve summarized some of his more interesting or entertaining points from both, along with some insight from Peter Christy, an analyst with 451 Research who also spoke at the seminar.
1. “2014 is going to make people cry.”
That’s Weckl’s prediction for switching revenues, and he thinks the industry is in denial about it: “I don’t think the market is ready for how bad pricing is going to be.”
He expects 10-Gb/s port pricing to drop 25 percent this year. Less than $100 per port for white-box switches “would be realistic,” he said.
2. The white box revolution is big. Yet, it’s not so wide.
White-box switching — the use of off-the-shelf hardware running the customer’s choice of software — essentially comes down to “just two customers,” Weckl said. They’re Google and Amazon — and Amazon didn’t hit the radar until last year.
Even so, white-box switching, if it were considered a single vendor, would be No. 2 top-of-rack seller behind Cisco, Weckl said.
That means a lot of business is concentrated at Google. And that’s a problem, because…
3. Hyperscale data-center owners are distorting the market.
“There are only going to be two folks who win Google for any given component,” Weckl said. “When you win or lose these cloud providers, it’s a significant share of the market that’s gone, and it’s going to take a significant amount of effort in medium or large enterprises to make that up.”
The implication is that being a market-share leader in the future means winning the heart of someone the size of Google, Amazon, or Facebook. Cisco certainly knows this; in fact, it’s because of “one cloud provider” that Cisco began using Broadcom chips in Nexus top-of-rack switches, Weckl said.
A similar effect is already happening in servers, Christy said. The cloud market is much smaller than the enterprise market, but it’s growing more rapidly. And Intel last year did see a shift in business toward ODMs and away from the likes of IBM and Dell, he said.
4. The market is leaking.
Cisco can defend its market share in medium-sized and large enterprises, but small enterprises are moving their networking business into the cloud, Weckl said.
That brings up a point that Brad Casemore, an analyst with IDC, discussed with me at Interop: Workloads are moving to the cloud — Amazon, Microsoft Azure, Google, and so on — and he thinks infrastructure sales will follow. The cloud vendors want this trend to accelerate; they’re vying for static workloads as well as dynamic ones, Casemore noted.
As for why Cisco can defend its turf in larger enterprises: “What Cisco has done is deliver working networks to customers,” and in most cases, a competitor needs to deliver “an equivalently good working network,” or at least produce significant opex reductions, Christy said. Of course, open-source switching is the antithesis of that approach, at least at this early stage. The market for the Cumulus and Pica8 is “very large, competent Linux shops,” he said.
5. China is stretching out of vendors’ reach.
Weckl pointed out that the maturing of Huawei and other vendors (ZTE comes to mind) is bringing about a situation China has been striving for: Chinese service providers can buy nearly all their equipment from Chinese vendors. “Ultimately, China is not a friendly market for the U.S., and that was going to happen regardless of politics,” he said.