“Fibre Channel over Ethernet is like a fast car,” said consultant Howard Goldstein of Howard Goldstein Accociates Thursday on his session about FCoE at Storage Decisions Toronto. “It looks great, but it probably won’t run as well as you thought or be as cheap as they say it’s going to be.”
Goldstein’s point basically boiled down to: if Ethernet’s good enough to be the transport layer, why bother layering an FC protocol on top of it? He dismissed the common answer to that question, which is that mixing FC and Ethernet will allow users to maintain existing investments in FC systems, saying it’s a myth. “You’re going to have to buy brand new HBAs and Fibre Channel switches to support FCoE,” he said. “Is this really the time to reinvest in Fibre Channel infrastructure?”
Instead, Goldstein pointed out that FC services such as log-in, address assignment and name server, to name a few, could be done in software. “Those services don’t have to be in the switch–Fibre Channel allows them in the server,” he said. He also questioned the need for a revamping of the Ethernet specification for “Data Center Ethernet” capabilities. “Is converged Ethernet a real requirement or a theoretical requirement?” he said. He also questioned whether or not storage traffic was really fundamentally different from network traffic.
However, users at the show said FCoE is still so new they weren’t sure whether or not to agree with Goldstein. “It’s too immature to say right now,” said Maple Leaf Foods enterprise support analyst Ricki Biala. He also pointed out an all-too-true fact: in the end, such technology decisions will be based on equal parts politics to technology. “It’s easier to convince management to buy in if you’re going the way the rest of the market’s going,” he said.
Your thoughts and quibbles on FCoE are welcome as always in the comments.