The growth of Fibre Channel over Ethernet
Fibre Channel over Ethernet is likely still a few years away from mainstream adoption; however, the Dell'Oro Group believes that this adoption will eventually happen. In the meantime, there are a number of hurdles that need to be overcome.
Ethernet has spread from its initial deployments in enterprise data networks to home and service provider networks and encompasses a broad array of applications, including voice and video. The latest arena for the potential broader adoption of this technology is within SANs as a possible replacement for the incumbent Fibre Channel (FC) technology.
Fibre Channel over Ethernet (FCoE) consists of an FC frame encapsulated inside Ethernet, which has been enhanced for converged data center deployments. This converged fabric provides three interrelated benefits for the end user. First, users could operate and manage their SAN and LAN as a single network, with native Ethernet connections. Second, since separate interfaces for FC and Ethernet would no longer be required, the number of ports per server could be reduced. This would lead, in turn, to a reduction in the number of switches required, as the separate switches currently used for FC and Ethernet traffic could be combined into a single FCoE switch. Third, enterprises could reap cost savings from combining storage and data networking administration teams.
FCoE is still a few years away from mainstream adoption; however, we believe that this adoption will eventually happen. In the meantime, there are a number of hurdles that need to be overcome. For example, FCoE is an immature and unproven technology, and SANs back up critical company data where there are often associated regulatory requirements. As touched upon briefly, FCoE uses a new class of Ethernet that has been enhanced to enable features already inherent in FC, such as lossless transmission and low latency.
Furthermore, FCoE currently carries a price premium not just over the fastest available FC speed (8 Gbps), but also over 10-Gbps Ethernet. Currently, on a like-for-like basis, FCoE products generally have a price premium in excess of 50% over non-FCoE-enabled 10-Gbps Ethernet products.
While technology and cost are significant hurdles to FCoE adoption, the “human factor” may be an even bigger one. SAN administrators and IT personnel are familiar and comfortable with the reliability and resiliency of FC.
Many end users we interviewed pointed to these challenges. Despite the potential for overall cost savings that FCoE offers (fewer adapters, switches, cabling, and potentially administrators), these users remain skeptical about its adoption. Much of their skepticism stems from FCoE being a new technology that they fear may not compare favorably with the quality, reliability, and peace of mind associated with the incumbent FC technology.
While we acknowledge the hesitation about FCoE adoption on the part of many end users, we are forecasting a strong ramp for the technology. We note that all four major SAN vendors have product offerings and major programs in place to spur FCoE adoption, and we expect that some of the nontraditional SAN vendors will also offer credible FCoE products. Additionally, a number of the storage integrators have already tested and qualified FCoE products. Finally, some of the end users that initially expressed skepticism about FCoE adoption have softened their stance and are now even evaluating FCoE deployments.
To some degree, this early end-user skepticism mirrors that which has surrounded past technology transitions, such as the adoption of voice over IP (VoIP). This leads us to believe that FCoE adoption is likely to happen, but the transition from FC to FCoE, like that of legacy digital voice to VoIP, may span the majority of a decade.
Figure 1 shows that after 10 years, VoIP is now only on the verge of becoming the majority of total PBX lines. However, we are mindful that data networking equipment (both FC and Ethernet) tends to have a shorter lifecycle than that of voice equipment‚ÄîPBXs generally are retired after seven years versus approximately four to five years for switches, and even fewer years for adapters. So the transition to FCoE may take less than that of VoIP. More specifically, we anticipate that by the end of 2013, more than one in every four combined FC and FCoE port shipments will be FCoE. The comparable adoption rate was one in every six for VoIP.
Since FCoE is currently available at one speed only (10 Gbps), its adoption depends on the adoption of 10-Gbps Ethernet. Ethernet at this speed currently is enjoying rapid growth, as it has become the technology of choice for the data center. It is already a very popular interface for data center aggregation switches, and it will likely become the interface of choice for server access and “top of rack” switches over the next few years. A plethora of recent product introductions into this space have brought low-cost and high-density 10-Gbps Ethernet ports to the server access layer.
In tandem with these releases, we are starting to see the adoption of 10-Gbps Ethernet on servers. This is largely due to the significant bandwidth discount versus lower-speed connections, since most servers do not as yet require this much bandwidth. While the attach rate of 10-Gbps Ethernet is not yet as high as what it was for 1-Gbps Ethernet (mostly due to the lack of a lower-power 10GBase-T transceiver), we are starting to see rapid growth in this area both on the adapter card side and even directly on the server motherboard. As a result, we expect that within the next five years, 10-Gbps Ethernet attach rates on servers will start to approach the rate that 1-Gbps Ethernet achieved after a similar time period (see Fig. 2).
So what does all this mean for the optical market? Like many other technology shifts, there will be both positive and negative effects.
On the negative side, the adoption of FCoE will result in fewer server and switch ports; hence, fewer optical pluggable modules and fiber cables will be needed. More specifically, where there was an FC server adapter connecting to an FC switch, and a separate Ethernet adapter connecting to an Ethernet switch FCoE can enable a single converged connection. Moreover, copper-based connections are very common for Ethernet; therefore, this shrinking number of connections will be split further between copper and fiber media.
On the positive side, we expect that FCoE will increase the percentage of servers connected to the SAN, which we believe currently hovers around 20%. Storage backup is one of the main applications that drives the need for bandwidth, and increasing demand for storage backup should lift the demand for all high-speed connections‚Äîcopper and fiber alike.
Furthermore, the combining of LAN and SAN traffic into a single pipe will increase that connection's utilization rate. In turn, a higher utilization rate should require higher-speed aggregation and uplink connections‚Äîi.e., 40 Gbps and 100 Gbps‚Äîthe majority of which will likely be optical.
FCoE on the horizon
FCoE is likely years away from widespread deployment. However, the benefits of running what once were separate networks as a single, unified network‚Äîwhile initially painful‚Äîwill be compelling. The transition to FCoE likely will be a mixed bag for the optical market, which will suffer a decrease in demand for total connections on the one hand and an increase in demand for higher-bandwidth interfaces on the other.
Seamus Crehan is vice president of the Dell'Oro Group (www.delloro.com).
Lightwave:The Business Case for Migrating to 8-Gbps Fibre Channel
Lightwave:Clouds on the Horizon
Lightwave:FTTE Battles for Enterprise/SAN Acceptance