Creating a Converged Data Center

Fibre Channel (FC) is a networking technology that was formally introduced in 1994 by the American National Standards Institute (ANSI). FC operates at various Gigabit speeds, typically using fiber-optic cable as a transmission medium (although copper cables can be used), and is used in Storage Area Networks (SAN). FC provides its own fabric and is a completely separate technology from Ethernet. Chances are that most people have heard the word “Ethernet” and have at least some idea of what it is used for (“computers or something, right?”). On the same note, it is also very likely that those same people have never even heard of “Fibre Channel” (“what the heck is that?”). This is because Ethernet is such a ubiquitous/widely-accepted technology that is integral to networks across the world and is here to stay. FC can be seen as “lagging” Ethernet in performance. While Ethernet has recently introduced 40/100Gb speeds and has Terabit Ethernet on the horizon, FC has just come out with 16Gb and is looking into 32Gb. FC is not without its benefits though. FC is: reliable, congestion free, and efficient. Wouldn't it be nice if there was some way to converge both of these technologies into one universal fabric?

Enter: Fibre Channel over Ethernet (FCoE)! FCoE is a relatively young technology introduced in 2009 by the ANSI's T11 Committee and outlined in their FC-BB-5 standard. FCoE's function is to tunnel/encapsulate a FC frame inside of an Ethernet MAC frame. This allows FC traffic to be sent as Ethernet traffic. It is important to note though that FCoE operates directly above Ethernet (not above TCP/IP) and therefore is not routable at the IP level (still confining it to the SAN).

Despite the benefits, there are a few issues to be aware of. As mentioned before, FC is a reliable and congestion free technology. In other words, FC is lossless and uses its own form of flow control (BB Credit). On the other hand, Ethernet can be seen as lossy and does not have priority-based flow control. To successfully implement FCoE, these issues need to be addressed.

Thankfully, the IEEE has introduced a set of standards referred to as Data Center Bridging (DCB). DCB includes four different technologies as follows: Priority-based Flow Control (PFC), Enhanced Transmission Selection (ETS), Congestion Notification (CN), and Data Center Bridging eXchange (DCBX). PFC is used to inhibit the transmission of frames on specified priorities. ETS is used to allocate bandwidth to certain priorities. CN is used to transmit congestion information end-to-end. DCBX is used by peers to exchange capabilities with one another. Each one of these technologies works independently from one another, but they all strive to meet the same goal: avoiding packet loss.

With DCB allowing FCoE to function properly, FC can now be sent as Ethernet traffic with minimal changes to the FC protocol or the frames themselves. Now, instead of there being two separate fabrics consuming power, money, resources, and making heat in the data center, there is one converged, universal fabric that produces the best of both worlds.

Matthew Goepfert, Fibre Channel Technician