“FCoE seems like a no brainer.” That was the opening line of a recent article on Enterprise Storage Forum. So why, the author asks, does FCoE seem to be struggling to catch on?

From the Xsigo perspective, the follow-on question is, “If FCoE is slow to grow, then why did Xsigo expand 3X in the past year?”

Although Xsigo does not use FCoE technology, the comparison is still relevant. FCoE is the technology that Cisco and a few others use to enable virtual I/O. Xsigo does not use FCoE (we use conventional Ethernet and InfiniBand), but the idea is the same.

So the question is, why is FCoE slow to launch, and why is Xsigo seeing different dynamics? Here are the points from the article, along with the Xsigo comparison.

Point 1 – Storage is slow to evolve: Enterprise storage managers tend to be risk averse (for good reason). Furthermore, when there is a new interconnect, interoperability needs to be proven on a wide range of gear. As a result, new technologies such as FCoE can be slow to catch on. Fibre Channel was launched in 1988 and took about 10 years to catch on. iSCSI came on the scene around 2000 and took at least 5 years to catch on. So it makes historical sense that FCoE would have a gradual adoption.

So why is Xsigo different? Xsigo presents a standard Fibre Channel interface. The SAN is connected to standard Qlogic FC silicon inside the Xsigo I/O Director. Within the server, the OS or hypervisor sees standard FC drivers. Because the technologies are familiar and everything behaves in an expected way,  user comfort grows quickly.

Point 2 – FCoE requires infrastructure change-out and upgrade: FCoE tends to be part of a larger transformation. New core devices, new types of servers, potentially even new storage. So it’s naturally going to be viewed as a long-term initiative.

Xsigo on the other hand is viewed as a top-of-rack unit, analogous to an aggregation device. The core does not change (the core ports are the same, you just need fewer of them), and you can use whatever blade or rack servers you want. The server just needs to have a standard 10G Ethernet NIC or widely available InfiniBand HCA). So Xsigo can be easily deployed whenever new server I/O is needed.  

Point 3 – CNAs are evolving: With FCoE the fear is that if you deploy a bunch of CNAs today, will they be obsolete tomorrow? A good question, but it’s irrelevant with Xsigo.

With Xsigo, the interface in the server is a standard part: Regular 10G Ethernet or InfiniBand cards. The I/O intelligence all resides on hot-swappable modules within the I/O Director. So obsolescence is not a concern.

Point 4 – FCoE itself is evolving: FCoE multi-hop capability was just added, finally facilitating greater fan-out.

Xsigo has offered multi-hop from day one. The solution has always scaled.

Point 5 – Who manages a converged environment? In data center management, the cultural barriers are real. When you converge storage and Ethernet, it does merge technologies that are often managed by separate functional groups within the organization.

Xsigo offers role-based access controls that allows storage and networking managers to maintain their familiar roles. The Xsigo hardware design is well-suited to this because the I/O intelligence for these functions is divided into physically separate modules.

The Enterprise Storage Forum article concludes that “FCoE is coming.” That is no doubt true. And, by the way, when Xsigo customers want to deploy FCoE-native storage, we will offer an FCoE module that will plug into the systems they are deploying today. So Xsigo again prevents obsolescence.

But Xsigo does not face the same issues that today slow FCoE adoption. Which is probably why Xsigo is seeing rapid growth and is in production deployment at hundreds of data centers worldwide.