A new Fibre Channel (FC) standard is in the works that leverages the venerable Ethernet architecture. Products based on the standard could appear in a few years, but it's unclear whether storage managers, particularly those who are happy with their FC environments, will be interested.
Last April, a collection of IT vendors (including Brocade, Cisco, EMC, Emulex, IBM, Intel, Nuova Systems, QLogic and Sun Microsystems) proposed the new Fibre Channel over Ethernet (FCoE) standard to the T11 Committee of the American National Standards Institute, which would allow SAN I/O traffic to be transported over Ethernet networks. Unlike protocols like FCIP, which encapsulates FC and sends it over a TCP socket, FCoE would map directly onto Ethernet.
According to Greg Schulz, senior analyst and founder of StorageIO Group, Stillwater, MN, a key benefit of FCoE is that it could potentially "remove the overhead of TCP/IP from the equation," leveraging the underlying Ethernet-based network while supporting FC frames and upper-level protocols.
To some degree, FCoE is certainly a push on the part of FC vendors to maintain the storage protocol's competitiveness once 10 Gigabit Ethernet (10GbE) firmly establishes itself in networked storage products. With 10GbE prices expected to drop around mid-2008, iSCSI adoption is expected to pick up, which could theoretically represent a serious threat to FC storage vendors.
The intent of FCoE, then, would be to let storage managers bridge their FC SANs to the rest of the data center environment, take advantage of the speed of 10GbE, deliver on the promise of a converged network and extend the life of their FC products.
The standard could have legs, "provided that the standard is baked and there's reasonably strict adherence to that standard ... so that things work together," says John Webster, principal IT advisor at Illuminata Group in Nashua, NH.
At first glance it would seem that FCoE would challenge iSCSI. But analysts stress it isn't a question of FCoE overrunning and eliminating iSCSI-based storage.
"The iSCSI debates will continue and iSCSI is not going away any time soon," says Schulz. "The only way FCoE takes out and eliminates iSCSI is it has to bring its value proposition and its cost effectiveness way down," he says, adding that it would probably be four years to six years before that could be achievable.
Some vendors and analysts believe the two technologies will become stratified, with iSCSI settling into the small- to medium-sized business market, and FCoE propagating at the enterprise level. Other observers, like Schulz, think it's more likely storage managers will want the technologies to coexist in different networks on the same wire.
But it's too soon to tell what the landscape will look like if and when FCoE-compliant products do arrive--conservative forecasts put the appearance of prototype technology around 2009, at the earliest--or whether iSCSI on 10GbE will be entrenched enough by then to dampen interest in FCoE.
By offering a way to interconnect FC networks across 10GbE to the rest of the data center, FCoE could extend the life of FC by as much as 10 years, say analysts. Diehard FC loyalists, however, may not be even looking at 10GbE. "They'll probably say, 'What do I need 10 Gigabit Ethernet for? I'll just stick with my 4 Gigabit Fibre Channel, thank you,'" says Marc Farley, director of customer programs for EqualLogic in Nashua, NH. "Manufacturers in the Fibre Channel world ... have a case here that says customers don't want to move off their Fibre Channel. The question is, in that case, 'Why do they want to go to this 10 Gigabit one, anyway?'" --Trina MacDonald