Emulex Heads Backwards At 80 MPH

Today Emulex announced the general availability of its "UCNA" to channel partners. However, I think it's important to be clear about what they have actually announced: they are making ServerEngines' 10GbE NIC available to channel partners. There are neither FCoE nor iSCSI capabilities on the card; it ships with Ethernet support. From all indications, it's a ServerEngines board with an Emulex logo.

Perhaps the most precarious part about today's announcement is Emulex's claim that it will sell these 10GbE adapters "now" and offer "field enablement options" to FCoE and iSCSI in the future as part of its "Pay-As-You-Go" strategy. But upgrading a NIC to support FCoE is a disruptive process that requires downtime. How do you upgrade an environment that has 600 adapters? This debatable strategy would require an army of field engineers to go out en masse and upgrade adapters while customers shut down their servers, or at best, cause path failure and induce performance degradation. Does Emulex plan to go out and rent the new Greyhound bus worldwide to dispatch legions of service technicians at customer sites with adapter upgrade toolkits in hand? It sure doesn't seem to me that upgrading Ethernet adapters directly at end-user sites with customer engineers is consistent with the established OEM go-to-market model for NICs and CNAs.

Furthermore, it seems to me that FC shops are going to FCoE, not pure 10Gb Ethernet. So Emulex's strategy of selling 10GbE adapters now and "upgrading" them later is confusing at best as no FC shop will go from FC to pure Ethernet and wait until Emulex can get them up and running on FCoE. Emulex seems to believe network convergence is about 10GbE before anything else; but it my opinion it's not. It's about FC. You know, the FC in FCoE. That's why Broadcom tried to buy the company in the first place--to get their hands on the FC stack.

Emulex's confusing route to Ethernet pits them head-to-head against Broadcom and Intel. It appears that what Emulex is trying to do is the equivalent of going backward at 80MPH in order to enter the Ethernet market space - hard to control and something has got to give under the stress. Emulex is leaping into the Ethernet world feet first - a world where they have little expertise or credibility. Emulex's confusing route into Ethernet is one that even Broadcom knew wouldn't work. The company has to cross two barriers to make its "UCNA" strategy work: first, Emulex has to somehow combat Broadcom and Intel to become a credible Ethernet supplier (this will be a tough row to hoe); secondly, they must attempt to become a credible FCoE supplier (see the diagram below).

Emulex is entering into a low-margin Ethernet business with NIC card that must be costly for them, as apparently it contains the underlying hardware to support FCoE and iSCSI at some point in the future. It's pretty obvious that Emulex is not going to make much profit margin on its "UCNA" NICs, since it appears they are sourced from ServerEngines. So do they plan to make up for this loss on their "FCoE upgrade strategy?" This will likely be costly to the buyer in the end. In my view, this is an unsustainable business model. Any time you're going from selling fully functional adapters with excellent margins to selling upgrades on off-the-shelf NICs, you're setting yourself up for troublesome times to come when reporting quarterly earnings.

If customers actually wanted an erector set-style CNA, Brocade, Broadcom and QLogic would have jumped at that business model to support their customers. It appears to me that Emulex is embracing a plan of necessity rather than a well considered strategy of choice.

I was thinking about some of the vendors in this space are moving in certain directions, after reading the Emulex announcement.Below is a diagram of how I see these vendors today.