But a couple of developments in the last few years have generated more interest in the neglected network. Fibre Channel over Ethernet (FCoE), touted as the great storage uniter, offers a way to knit together otherwise disparate networks, making Fibre Channel and Ethernet networks function as one (at least in theory).
FC has been the de facto standard connective tissue for enterprise-class storage environments, and with its storage-centric protocol design it's generally accepted as the performance leader. But Ethernet's stake in the storage shop has grown significantly over the last five or so years as NAS systems have proliferated in response to the ever-rising tide of file storage. Ethernet's reach has also been extended as iSCSI continues to gain a share of the block storage market. iSCSI didn't exactly take the data center by storm, but based on our research, its incremental growth puts it in more than 40% of the country's data storage shops.
That makes Ethernet-based storage a player for both file and block. And it almost makes it the default candidate for multiprotocol storage, where block and file share the same system, which is quite possibly the fastest growing segment of the storage system market.
Every IT shop has years of Ethernet experience, and the hardware it hangs on is tried, true and pretty cheap. Fibre Channel aficionados will say that's all true, but you don't get the performance of a network specifically designed for storage. Well, now you do get that performance (or nearly so) as 10 Gigabit Ethernet (10 GbE) gear begins to replace the old 1 GbE stuff in most shops. Even at a slightly slower 8 Gbps, Fibre Channel might still have the performance edge because of its storage pedigree. But it's not that much of an edge, and the Ethernet roadmap, bolstered by the ratification of the IEEE 802.3ba standard last year, has 40 GbE on the near horizon with products already being demonstrated, and 100 GbE to follow hard on its heels.
Add the emergence of SAS disks as a replacement for FC disks on the back end and how SAS can easily combine with SATA drives for tiered multiprotocol storage, and it looks like the NAS/iSCSI/Ethernet combo will be pretty tough to top.
Where does that leave Fibre Channel over Ethernet? If Fibre Channel is getting squeezed out of the storage picture and Ethernet is destined to be the data center-wide network of choice, why should we even bother with FCoE? FCoE appears to be a bridge technology (pardon the pun) that does a good job of linking FC and Ethernet. The question is how long will we need that bridge, and how many companies need to connect those two environments? Is it worth investing in CNAs and special switches and new interfaces? We're not seeing a ton of FCoE storage systems, so maybe storage vendors don't see a lot of potential for a big payback on that investment.
Of course, there are some shops that can use FCoE now, use it effectively and save money in doing so. But I don't expect they're in the majority.
Ultimately, it looks like it'll be an Ethernet world, which will certainly make things simpler. And as I/O virtualization catches up with server and storage virtualization, working with a single network architecture will make it even easier to virtualize and share network resources, and to use them more efficiently.
BIO: Rich Castagna (firstname.lastname@example.org) is editorial director of the Storage Media Group.
* Click here for a sneak peek at what's coming up in the April 2011 issue.