Two weeks ago I was on a panel that discussed the future of Ethernet at the Ethernet Summit in San Jose.
It was a very interesting group of panelists, each presenting a vision of Ethernet and where it could go and better yet, what was needed to get there. One very interesting contrarian view that was presented was essentially around where Ethernet would not go. Several different areas such as broadcast TV and storage were highlighted as areas where Ethernet will not dominate. The latter one i.e. storage, took me a little by surprise. The presenter made the argument that fiber channel will continue to dominate as the storage technology of choice, and one that Ethernet would not be able to shake off.
Talk about Déjà vu. This reminded me of the arguments of the last two decades… Ethernet would not displace ATM, Ethernet would not displace voice PBXs, and so on. Sure it took time, but Ethernet endured and endeared – and in the end – prevailed.
Is the same going to happen to storage and particularly fiber channel based SANs?
Let’s look at both sides.
Fiber channel: Until recently, Fiber channel at 8Gbs did provide a pretty good leg up over Ethernet at 1 Gbps. Even prior to 8GBps fiber channel, Fiber channel at 4Gbps or even 2Gbps, provided a significant speed advantage over Gigabit Ethernet.
The connection oriented nature, along with the credit based flow control also did make fiber channel more predictable, more lossless, than “best effort” Ethernet.
As a result fiber channel became the technology of choice when it came to high performance enterprise grade SANs with few viable options to compete against it. Consequently, significant investment into tools, applications and services were made over the last decade or more, that allowed fiber channel to firmly entrench itself in the SAN space. Knowledge bases and expertise in fiber channel has grown over the years as well.
However, Fiber channel, and fiber channel switching in particular, is a lot less “open” or interoperable than Ethernet. Fiber channel deployments are tightly governed by a compatibility matrix all the way from fiber channel HBAs to fiber channel switches to targets. Indeed, the the fiber channel switching market is tightly controlled by a very small number of vendors making it almost a duopoly or an oligopoly at best.
Ethernet: On the other hand, the recent growth and adoption of 10Gbps Ethernet has now allowed it to overcome the speed advantage and argument of 8Gbps fiber channel. And with Ethernet moving towards 40Gbps in the data center, the argument of 16Gbps fiber channel pales in comparison. With 10Gbps Ethernet LOM expected later this year, the price performance curve of Ethernet is beginning to look very attractive and is forcing the industry to take a look at Ethernet as a viable alternative fabric of choice for storage networking.
The fact that Ethernet switching is far more open and interoperable compared to fiber channel also means that customers have more choices when it comes to vendors. The recent standardization work in the IEEE of Data Center Bridging standards (DCB) along with the work going on in the T11 committee under INCITS on FCoE (Fiber Channel over Ethernet) provides further impetus to the movement of moving storage networking to an Ethernet fabric. This is not to say, however, that FCoE will replace FC or that FCoE will become the technology of choice for SANs.
Rather, that with the economics of Ethernet now working firmly in its favor, more innovation in storage and SANs will now center around Ethernet based solutions. The recent rapid rise and growth of iSCSI which operates over TCP/IP and Ethernet, stands testament to the fact that Ethernet based storage is now enjoying a resurgence and is attracting more attention with renewed focus. Indeed iSCSI is quickly moving up the food chain towards becoming an enterprise class solution helped in large part by 10Gbps Ethernet targets becoming available. With more investment in tools, applications and platforms for Ethernet based storage will come higher performance, better predictability and more reliability. Just as importantly, the open nature of Ethernet will encourage more competition and ultimately more choices for customer.
However, the last decade of fiber channel dominance does mean that a large number of enterprises do have an ongoing significant fiber channel investment. Both from a capital perspective, as well as in terms of tools, training, and expertise around their fiber channel deployments. As a result one can certainly expect that there will continue to be a period of time where fiber channel will continue to be the mainstay of enterprise storage, albeit as islands of Ethernet based storage infrastructure start getting built out around it. Already, technologies to “bridge” fiber channel to Ethernet are beginning to surface.
Now if I were a betting man, I would say that that is the beginning of Ethernet taking its place in the enterprise storage realm.