7/8/2008 -- This spring, Cisco Systems Inc. announced its Nexus 5000 series of Server Access Switches. Cisco's go-to-market strategy with the Nexus 5000 involves a pair of deployment options: It can be used as either a central switch in a Fibre Channel-over-Ethernet (FCoE) fabric -- which would give the company a compelling next-gen datacenter pitch -- or in conjunction with its 7000 Series Aggregation Layer switches. In the latter schema, Cisco said, the 7000 serves as an aggregator of traffic from several 5000 Series switches.

That's the backstory. What's intriguing, according to industry watchers, is that Cisco's Nexus 5000 Server Access Switch is one of several new switching deliverables that (both individually and collectively) augur a next-gen convergence -- in this case, of two venerable storage networking topologies, SAN and NAS.

It's a no-brainer combination that's been years in the making, according to industry veteran John Webster, a principal IT advisor with consultancy Illuminata. If Ethernet was fine for NAS, why couldn't it suffice for SANs?

"Networked storage was introduced in the early 1990s as Network Attached Storage [NAS] and Storage Area Networks [SAN]. One of the advantages of NAS was that it used existing networking infrastructure -- Ethernet and IP. On the other hand, SAN offered scalability, but required the installation of a new network based on the Fibre Channel [FC] protocol," Webster explained.

This had some CIOs scratching their heads.

"When confronted with SAN project proposals, CIOs often wondered why they had to install another entirely different network based on FC to get the scalability benefits of networked storage," Webster said. "Can't we just have one network to manage? Wouldn't that be easier and more cost-effective for all involved?"

As it turned out, those head-scratching CIOs were on to something, Webster said. Now the networking industry seems ready to hop on the bandwagon.

For starters, according to Webster, there are a couple of convergence-precipitating technology standards -- "lossless" Ethernet and, of course, FCoE.

But there are also a bevy of hard technology solutions -- "converged" switches from Cisco, rival Emulex Corp. and others. Take Cisco's first Nexus 5000 Series deliverable, the 5020. It's a 2U, rack-mounted switch that's populated with 40 fixed 10 GbE/FCoE ports; but up to 16 additional ports -- of FC, FCOE or some combination thereof -- can be added, according to Cisco.

Experts envision any of several converged deployment scenarios for Cisco's 5000 Series offerings. For starters, Webster said, customers can tap them to connect existing Ethernet-connected assets to FC fabric-connected assets. Once FCoE SAN storage becomes feasible, Webster continued, they can convert to an all-Ethernet backbone, if they want.

Secondly, there's an embrace-and-extend deployment opportunity: rolling out new blade servers (or adding existing blades) to a SAN fabric. In this scheme, Webster said, all of the blades would use FCoE. Finally, there are a number of greenfield opportunities, too, such as new datacenter or application deployments.

The move to FCoE will require increased coordination between and among IT groups, Webster warned. "IT administrators will see at least one major difference in operational management. With FCoE, more cross-coordination between server teams, networking teams and storage teams may be required," he said. "Where server teams manage both the HBAs and the NICs, they will likely manage CNAs and will continue to take direction from the storage team as far as what is a qualified and supported FC HBA configuration in terms of driver, firmware versions; parameter settings, etc."

There's a sense in which FCoE triumphalists might be missing the point. Because if FCoE does emerge as the converged datacenter backbone of choice -- and the still-extant iSCSI standard might have something to say about that -- it'll do so largely because of its Ethernet component.

It's possible, then, that high-bandwidth Ethernet (e.g., 10 GbE and above) might play spoiler to FCoE's coronation.

"The more momentous question is whether or not high-bandwidth Ethernet will replace FC as a datacenter I/O transport. Possibly, but the process will take years -- many years," Webster said. "It took a decade to get to where we are now with SANs, and a fair amount of storage is still directly attached to servers, even in large datacenters. IT administrators are notoriously risk-averse when it comes to the integrity and reliability of the server-storage connection. So we can expect a lengthy process of testing and vetting before FC-based networked storage is replaced by Ethernet." --Stephen Swoyer