This year’s Storage Networking World Europe show focused attention on Fibre Channel over Ethernet (FCoE). However, while SAN vendors promoted FCoE as a way to extend Fibre Channel’s reach and longevity, skeptics described it as unnecessary — and perhaps even a Cisco-powered Trojan horse intended to wipe out rival networking supplier Brocade Communications.
FCoE is a draft specification for carrying Fibre Channel storage traffic over 10G bps (bits per second) Ethernet cabling. Unlike iSCSI, a rival scheme for carrying storage traffic over Ethernet, FCoE leverages today’s Fibre Channel SANs (Storage Area Networks) and administrative tools, so it looks very familiar to storage administrators.
It also needs a single network card instead of two — one for Ethernet data networking, and another (called an HBA, or Host Bus Adapter) for Fibre Channel — explained Joe Gervais, senior product marketing director at SAN specialist Emulex.
“We call these Converged Network Adapters, or CNAs, because they are a superset,” he said. “One card means customers can use physically-smaller servers and a single network switch, but the SAN infrastructure looks the same as before.
“It enables Fibre Channel to go further,” he added, arguing that organizations can use it to connect additional servers to their Fibre Channel SANs without the cost of more Fibre Channel hardware.
“I don’t get it,” retorted Steve Duplessie, senior analyst at the Enterprise Strategy Group. “Why do we need it? If you want to connect a new server to Fibre Channel, put an HBA in. Why do you need a new card? The server will already have Ethernet.”
Even FCoE’s big fans — who include CNA suppliers Emulex and QLogic, Cisco, and NetApp, which has developed the first FCoE-capable storage server — say that it will appeal only to existing Fibre Channel customers.
It may not even do that, warned Brocade solutioneer A.J. Casamento. He pointed out that FCoE is not exactly cheap — along with new CNAs, it requires new switches capable of supporting Data Center Ethernet (DCE). This adds bandwidth prioritization and congestion management protocols to make Ethernet lossless, meaning that data packets will never be lost.
“There is always the potential for cost savings if we can drop an infrastructure,” he said. “Near-term though, there’s the cost of the [FCoE] componentry — and by the way, did anyone mention you also need new cable?”
That’s because DCE, at 10G bps, won’t run on most of today’s 100M bps or Gigabit Ethernet cabling, he said.
There was skepticism too over Cisco’s motives. Cisco recently announced the first commercial FCoE-capable switch, the Nexus 5020 (Brocade demonstrated an FCoE switch in Frankfurt, but said it would not announce products until after the specification is approved, probably in 2009) but its interests are far broader.
For example, lossless DCE is not just for FCoE — it will also be very useful for NAS (network-attached storage) and iSCSI, said Cisco consulting systems engineer Ulrich Hamm.
Steve Duplessie suggested that, whether FCoE hangs around as an Ethernet-based protocol or not, Cisco’s long term goal is to remove Fibre Channel as a hardware infrastructure. “Cisco wins by no one needing Brocade any more,” he said.