Tom Petrocelli's take on technology. Tom is the author of the book "Data Protection and Information Lifecycle Management" and a natural technology curmudgeon. This blog represents only my own views and not those of my employer, Enterprise Strategy Group. Frankly, mine are more amusing.

Thursday, June 25, 2009

FCoE is Rubbish

I'm beginning to understand the debate about the Fiber Channel over Ethernet (FcoE) concept and I don't like it. At first it made little sense to me. iSCSI delivered most of what we needed in terms of cheap SANs. It leveraged the IP infrastructure already in place in just about every organization in the world. iSCSI also made use of 40 years of experience, knowledge, and training. For high performance you went with Fiber Channel since iSCSI couldn't meet the very high loads of some systems. Otherwise, iSCSI was good enough for a lot of applications.

So why FCoE? The cost structure is likely to be higher than iSCSI or at best the same. The performance wouldn't be the same as pure FC. The idea that you need a gateway to the IP network doesn't really make sense to me either. Who really does that anyway? A few folks doing long distance SANs perhaps but there are tools for that and cost is not a big problem in those environments. You can always gateway to iSCSI if needed and the hardware for that already exists.

Then it occurred to me. It's marketing, stupid! When networking vendors such as Cisco go out and sell their SAN products, they are generally on the same footing as the FC players like Brocade. Brocade has more experience and knowledge about SANs which translates into an advantage for them. They also know the storage folks inside large companies that network vendors don't. Those storage guys like having their own flavor of network technology. It keeps the network admins out of their shorts.

Now, FCoE starts to make sense. You can sell IT on convergence or native integration or unified platforms or whatever marketing babble you choose. It is hoped that management will rally around the idea of having one networking platform even for two different types of network applications. Don't kid yourself, SANs and LANs are very different network applications with very different technology needs.

Best of all, as a networking vendor you have the upper hand in the sale. You can insert your champions (the core networking folks) into the process. You can sell expertise that the FC guy doesn't have. The worst case scenario has you on equal footing to the FC vendor where you can sell on the merits of the products. Of course, with 30 more years of experience in Ethernet, you will have a few tricks up your sleeve that the FC guy doesn't. Nice position to be in.

FCoE as a convergence/integration/unified platform play is rubbish. No one is going to run SAN traffic and LAN traffic over the same Ethernet network. It will still be two pools of equipment, much of it specialized to FCoE. Most of the real networking expertise in a company is in the IP space so no real advantage there. Once you start to install the specialized FCoE switches and NICs (or brand new unified platforms using a forklift I imagine) the costs won't be that much different.

iSCSI makes sense in so far as it provides a low cost SAN option for low to mid-performance SANs. Old fashioned FC makes sense because it provides a proven high performance storage networking capability for intense applications. FCoE does neither. All it does is give networking vendors a leg up against existing FC vendors.

Great marketing I must admit. Not convincing technology but a good way to position SANs as just another network flavor. If there is a technology advantage here it doesn't seem to create much of a business advantage for the IT folks.

I'm sure I'll get a bunch of hate mail telling me all the minor advantages of FCoE, many from corporate mouthpieces. Instead of wasting your time on that, tell me why someone will pay money for FCoE rather than iSCSI or FC. Tell me why we need to gateway SANs from FC to Ethernet which can't be routed and hence not good for wide area applications like remote backup.

Don't go down the road of the unified platforms either. Saying you need FCoE to create unified networks is not true and, at best, self serving. Unified platforms happen because of software not Ethernet.

Just don't say convergence, okay.


Eugene said...

Very much agree with that take on FCoE, Tom (one of the few topics we didn't touch on in our face-to-face the other day). If low cost 10GBase-T comes along soon and provides an easy upgrade path for iSCSI SANs at far less cost and complexity than FCoE then FCoE's real colors will start to show. Gene

miked30000 said...

FCoE doesn't compete with iSCSI. It provides an evolutionary path for FC users, mostly in the data center, to adopt a single infrastructure. Why? Strong benefits at the server-facing network edge. In the short term it has a lot to do with eliminating cables and makes a lot of sense if the network side is already planning a move to 10GE. After all, most FC I/O is challenged to drive 2G at the server. It also seems to have a lot of merit in simplifying blade server architectures.

BTW, you don't give credit to Cisco who has a strong SAN product line in the data center and storage vendor relations.

Tom Petrocelli said...

In essence you are saying that it makes certain products easier to build like blade servers and unified platforms. Since most FC is up to 8Gb already, introducing yet another network into the SAN backbone doesn't make sense to me.
I am also well aware of Cisco's SAN products having helped to develop and deliver the first sales training for MDS. Given that they have FC products, the move to FCoE seems unusual, perhaps undermining their own existing products.
BTW, I almost didn't post this since you have no profile on Blogger and no other identifying info. I get a lot of corporate mouthpieces that spout preprogrammed garbage. At least you made a cogent argument which I liked. In the future, identify yourself so that you can take credit for your opinions. I can respect them even when I don't agree with them.

J. R. Blosil said...

Interesting article. I don't subscribe to the idea that IT managers will resist combining networks for data and storage. Notice the trend in VoIP over shared networks. That may have been considered risky a few years ago. Economics are driving need for efficiency. And if the solution meets the needs (including security and performance) and delivers increased benefits, then the technology will be considered.

Tom Petrocelli said...

I think they will combine networks and it's called iSCSI. I question why we need yet another storage protocol that runs predominantly over Ethernet. iSCSI may not have to run over Ethernet (could run over a WAN theoretically) but it almost always does.