Standards come and go, and some never have lift off; but I think FCoE is one that will ramp up—as evidently does QLogic. The starting point is that the FCoE standard is here, backed by some products, and has been tried and tested. It is now beginning to ship.
What does it do? It does what it says on the tin—sends FC traffic over Ethernet cables alongside the Ethernet traffic. This is an acknowledgment that FC is not going away any time soon because what it does not do is any convergence of FC format to, say, IP (but Ethernet will deal with IP traffic as well).
So how does it help? Major reasons why FCoE will grow are cost and flexibility—and specifically for the larger organisations carrying both an FC network and an Ethernet one. It is not a matter of the FCoE converged adaptors being low cost—they certainly cost more than Ethernet NICs if not FC HBAs—but that using the technology will allow a simplification of an organisation's physical networking infrastructure.
Perhaps the biggest deal in this is that a single set of cabling to cover both network types suddenly becomes practical. Existing cabling costs may not immediately be reduced because that money has already been spent; but when a decision to extend or change the network occurs, the user knows that 10Gb Ethernet copper cabling with converged adaptors covers most future eventualities—and costly FC cabling is rendered obsolete.
The adaptor uses a single server port to handle both 10Gb Ethernet and FC HBA traffic, so this potentially frees up plenty of comms slots on each server; then the mixed traffic behind it will share one cable instead of needing two. There is also a potential for serious power savings in this.
Tell-tale signs of take-up include IBM and HP both announcing FCoE in the last couple of weeks. Scott Genereux, senior VP of worldwide sales at FCoE adaptor provider QLogic, was very bullish when I spoke to him last week, when he said: "We see every major storage vendor supporting FCoE by the end of the calendar year." (Of course, if this fails to materialise as expected then the scope for infrastructure savings will be considerably reduced.)
His view was that a change of attitude began when QLogic succeeded in shrinking the technology onto its single chip converged network adaptor (CNA). He expects QLogic to be the major beneficiary of the early takeup because, as I reported in April, it has a head start with its 8100 Series CNAs. These combine 10Gb Ethernet with FC at the full 10Gb speed (so above 8Gb FC) partly achieved through the in-built FCoE offload engine.
Genereux said IBM was going with QLogic, although the situation with HP was not as clear.
It is in the data centre that Genereux sees the biggest plus. "One person can manage both FC and Ethernet instead of two," he said. However, this will only be entirely true if the whole FC infrastructure is taken across to FCoE—meaning an upfront cost before the expected ongoing savings. In that situation QLogic Ethernet users will find the common APIs and management tools familiar.
Meanwhile he suspected that there was a higher comfort factor over the security of FC than IP on Ethernet, an argument for not converging the formats at this time.
QLogic has certainly made a major commitment and has begun shipping its CNA to the channel. (In EMEA volume shipments are due to start by the end of next month.) Genereux said this was backed by training for its channel partners, not least because they needed to understand and explain to end users where the value of such a switchover would come from.
Some will see FCoE as a short-term move. On the other hand, the death of FC itself has been forecast for a while and I see little sign of this. FCoE might actually prove a useful half-way house; once an Ethernet infrastructure has completely replaced FC, then the full move away from FC will become a less daunting task.
Who knows, anyway, if a completely new mega-advanced standard, needing different cabling again, might emerge in due course. If everyone, for some reason, went for it, it would still be easier to convert from one network than two. So the lack of convergence right now is not, to me, a huge deal.
Call me a cynic, but I would also not be surprised if some of QLogic's competitors (for instance FC-lovers Brocade and Cisco or arch-competitor Emulex) were playing down FCoE's potential—simply because QLogic has a technology lead right now.
Not that QLogic is so fool-hardy as to put all its eggs in one basket. Genereux realises iSCSI will be more popular at the SMB end; it has its own solutions there and recently purchased Netxan which supplies an Ethernet NIC with an IP offload engine. The company is also working hard with Infiniband for high performance server clusters (a market in which it has reached the number two supplier spot).
There is no denying QLogic is also considering a 16Gb FC HBA; if the demand proves to be there that only demonstrates greater FC longevity than some predict. By the same token, Ethernet could go to 40 or 100Gb.
Anyway I am not a betting man, but I think FCoE will do pretty well in the next few years—starting about now.