Fibre Channel over Ethernet (FCoE) is making its push into enterprise data centers, guided there by industry heavyweights including Cisco, HP and Brocade. As with any new technology, there is marketing hype both for and against FCoE data center networks. Enterprises will have to cut through all of the hype to determine whether FCoE and the whole concept of data center network convergence is the right path to follow.
Myth: Fibre Channel over Ethernet is the ultimate storage technology and enterprises should be planning deployment, if not already rolling it out.
Reality: FCoE, and the promise of converged data center networks, is still in a very nascent stage. Many products have been announced by Cisco, Brocade and HP, among others, but only subsets of those lines are currently shipping.
What may really work: Unless your organization has a compelling reason to jump on the converged network and FCoE bandwagon, a wait-and-see approach is recommended by most people who follow the market. In the meantime, it's important to keep an eye on the ongoing development of other network convergence technologies, such as Data Center Bridging (DCB) and even iSCSI. A converged network is a fundamental change not only to the data center infrastructure but to the IT organization as well. Careful consideration of the impact on both the storage and data networks will be needed. In addition, network and server inventories must be taken, and even management responsibilities will have to be reassessed in the move to converged data center networks.
Myth: FCoE is a stopgap measure, designed just to keep storage vendors relevant, and it will most likely be displaced by technologies like iSCSI.
Reality: The reality of this idea ultimately depends on who you are asking -- a storage specialist or a network administrator. Storage experts will tell you that FCoE is the next evolution of Fibre Channel, delivering on the promise of higher-performance storage by combining Fibre Channel with 40 G and 100 G Ethernet. Many networking pros, on the other hand, see FCoE as a means of protecting the investment in existing Fibre Channel networks, which will ultimately be eliminated as data center networks converge on the Ethernet network and TCP/IP stacks they know and trust.
What may really work: There are actually very few technologies that "kill" another technology. In the case of FCoE vs. iSCSI, each technology will no doubt find its place in enterprise networks, based on the particular needs and priorities of the organization. Some enterprises will capitalize on FCoE to maintain and evolve their existing Fibre Channel installations. Similarly, iSCSI will continue to find its way into enterprises, particularly those without an FC heritage that want a storage platform more akin to the rest of their data networks. Of course, there will always be a third kind of organization: those that have no interest in converging storage and data networks, instead maintaining their existing disparate networks.
Myth: Adding FCoE will slash our physical port counts and the amount of gear in our server racks.
Reality: Maybe. While it is true that moving to a converged network will eliminate the need to have both IP and storage network switches in the rack, there is still debate on whether even 10 Gb converged network adapters (CNAs) in the server will handle both a heavy storage and IP traffic load.
What may really work: A converged network will ultimately cut down on the physical connections and the need for as many ports on data center servers, but the massive reductions of port counts touted might be a bit extreme. I/O-intensive applications, such as Microsoft Exchange or large databases, have the potential to consume the lion's share of the 10 GbE available on a CNA sending data to the storage network. While the remaining bandwidth is still twice that of the typical 1 Gb network interface card (NIC), you will still need to consider the network and storage utilization of a given application to apportion the necessary bandwidth and port count for a server on a converged network.