For nearly ten years, storage and virtualization analysts who were supposedly in the know, predicted the imminent demise of Fibre Channel, to be displaced by protocols such as iSCSI or FCoE riding atop 10G Ethernet. Given the splash many storage vendors have made over 16G Fibre Channel, the ubiquity of Ethernet is nothing to anticipate in the next few quarters. So what happened?
Ethernet's dominance was hit by the double-whammy of continued high transceiver pricing that slowed the acceptance of 10-Gbit replacements for Gigabit Ethernet, and the brutal recession of 2008-09, which slowed new deployments of all networking and data center equipment. At the same time, cloud services arose in a manner that assured the expansion of existing storage clusters based on Fibre Channel, and even InfiniBand in some cases. Few could have predicted the extension of InfiniBand to 40 Gbits/sec, though the shrinkage of that market to two major vendors probably assures an eventual sunset of that technology.
There are plenty of reasons Fibre Channel may prove worthy through 32 Gbits/sec and faster speeds. Several large vendors, including Brocade and Emulex, help drive down the price of HBAs, and the per-port price of switches. The efficiency of storage transport using Fibre Channel allows it to serve as a key element of "green data center" initiatives (this is realized by moving to more energy-efficient protocols, such as 64b/66b encoding). And the continued development of FCoE assures the future peaceful coexistence of Fibre Channel and Ethernet.
Last fall, the SAN Equipment report of Infonetics Research predicted an aggregate $41.7 billion market for Fibre Channel switches, HBAs, and interconnect bridges between 2012 and 2016, if FCoE is considered part of the Fibre Channel market. Single-year sales of SAN Fibre Channel/FCoE equipment could top $9 billion by 2017, Infonetics predicted. At the time, Infonetics said Brocade owned 99 percent of 16-Gbit port shipments, though the advent of new players entering the market in 2013 would accelerate 16G growth beyond the 95 percent quarter-by-quarter sequential growth experienced in 2012. Dell'Oro has predicted that the crossover point when 16G exceeds 8G shipments was slated to happen in mid-2013, and may have taken place already. The important factor of market broadening in 2013 is not only the arrival of system-oriented companies such as Dell, IBM, Cisco, and Juniper, but the commitment of switch and HBA companies such as Emulex, Chelsio, and Mellanox to a market where Brocade holds a commanding lead.
There is a reason Brocade has joined a recent effort to rebrand 16G FC as "Gen 5." In LAN and WAN channels, speed is the primary factor in choosing a network backbone, and most physical-layer protocols such as Ethernet and Optical Transport Network follow powers-of-ten rules. Fibre Channel bears more resemblance to PCI Express and its Generations 2 and 3, in that users are interested in several software protocol features beyond the simple network speed.
In the early days of the enterprise data center, Fibre Channel was seen as virtually synonymous with SAN or network-attached storage. As new and complex layers of hybrid clouds emerge, more unique architectures with local multiplexing and hierarchical strategies of virtualization have emerged, giving rise to innovative architectures such as next Gen 5 cards for WDM-based networks. New topologies for Virtual Desktop Infrastructure and Inter-Switch Links likely will emerge as cloud infrastructure grows in layering and complexity. The turn in some advanced storage clusters from disks to solid-state drives virtually mandates the use of higher-speed Fibre Channel links.
The fact that INCITS T11 has a 32G Fibre Channel standard ready for 2014 draft suggests that technical standards bodies do not see any rapid demise for Fibre Channel – though T11 is keeping options open for further Fibre Channel or FCoE speed upgrades, based on market demand. The 28G serial baud rate of 32G Fibre Channel maps well into the 25-Gbit channel rate of 100G Ethernet and to a proposed 25G lane option for InfiniBand. In addition, the next round of Fibre Channel (Gen 6?), will be able to use the same SFP+ modules used by Gen 5 ports.
Five years ago, some analysts were predicting that iSCSI would replace virtually all Fibre Channel applications where native Ethernet was not warranted. The mid-range cloud market for iSCSI has been aided by the protocol's support for 10G Ethernet, though most network and storage analysts see iSCSI as having a more limited future than any competing protocol. Some enterprise networking analysts were anticipating a one-to-one battle between iSCSI and FCoE; now, a continued role for Fibre Channel Gens 5 and 6 appears likely, while iSCSI's popularity may be limited to small and medium businesses.
Infonetics and Dell'Oro both noted a slight decline in Fibre Channel sales in the first quarter of 2013, but both research groups attributed this to customers waiting for a true multi-vendor Gen 5 environment to emerge. Now that multiple Gen 5 products are available at switch, interconnect, and HBA levels, market analysts are anticipating very healthy growth for Gen 5. At least one additional generation of native Fibre Channel is likely to see a significant, dedicated customer base, before a converged FCoE market emerges. Few might have predicted this outcome in the mid-2000s, but few were anticipating the rich and complex cloud storage and server-cluster topologies that have emerged in the intervening years. By the end of the decade, this market may be served by a converged realm of Fibre Channel, Interlaken, and 100G Ethernet. In the meantime, Fibre Channel will enjoy several years of growth at 16G and 32G speeds.