Sea. (Converged Enhanced Ethernet)

While it hasn't been covered in any great detail on storage blogs I did get a question about FC over Ethernet so it might be time to take a quick pass at it.

What we're really talking about when we say FC over Ethernet is FC over CEE (Pronounced Sea), or Converged Enhanced Ethernet. CEE is an effort to extend the Ethernet spec to add new functionality and resiliency so it can be used for forms of traffic beyond what it's used for today. The ultimate aim of what we're talking about here is the ability to collapse three or more different types of Fabric (LAN, SAN, and IPC) into a single Fabric.

What's driving this? Well 10GigE for a start. Even though it costs a couple of grand per port at the moment that number will start dropping like a stone as time moves on, and without CEE there will be this high bandwidth infrastructure which is only useful for specific forms of traffic. Nuts to that. Now's the time to begin making modifications so that we can get as much of our other traffic on 10GigE as is possible.

I'm sure people are wondering why bother with FC over Converged Enhanced Ethernet when we already have things like iSCSI or InfiniBand. Lets face facts InfiniBand required people to start from scratch, no problem if you're starting from scratch as it is but a big problem if it's a case of rip and replace. IB has been a non-starter for a while and it could be the biggest loser if customers begin to adopt CEE.

Okay, but iSCSI is finally taking off why not just stick with that? The problem with iSCSI is that if you have an FC SAN there's probably a lot of FC management tools and hardware you've invested in which just isn't applicable in iSCSI's Ethernet only world. FC and iSCSI don't "run together", they run in parallel to each other and tend to be managed as such. So here comes FCoCEE to give your FC SAN the ability to extend out onto Ethernet which has been enhanced with features as to negate the best effort delivery nature of Ethernet and allow for traffic differentiation, congestion control, and so on. We're not talking about FC-IP either that's a long distance solution and there isn't the overlap you think there might be as FCoCEE is a single subnet solution, your data center being that subnet. When you'd jump subnets or shoot out across the Internet is where FC-IP would come in.

So how far along are we? In reality probably further than we all expect. Patents have been filed, silicon has been cut and chances are that trademarks are being dreamed up as I'm typing this. I don't expect this to get stuck in committee hell as a lot of the principals want to see this on the starting line as soon as possible. 10GigE appears to be the green light so they'll want things ready for when the price normalizes.

To conclude our very quick look at FCoCEE, the naysayers looked at FC's recent inactivity and declared it to be dead. In reality it was cocooned and in the process of evolving.      

Comments

[Mark]

I see the bit about FC management tools often stated as an advantage for FCoCEE, but no one ever gives an example of such tools. I have heard things like multipathing (i.e., EMC PowerPath) and replication (i.e., EMC SRDF). But it would likely be trivial to multipath iSCSI (or you could rely on spanning tree), and as for replication, that is almost always IP-based at some point (i.e., FCoIP).

So what are these management tools? Fabric management? Certainly IP over Ethernet has plenty of those available.

FC Zoning? That would be replaced by Layer 2 VLANs in FCoCEE.

iSCSI has really been gaining steam over the last couple of years. It is very popular in Microsoft Windows networks, and is gaining more interest in Linux. It is especially popular in multiple subnet environments (distributed servers, desktops, etc.), as it is routable at the IP layer. So iSCSI will not go away, it will only get stronger.

As iSCSI will remain with us, and only grow, the question is will customers consider a third protocol (FCoCEE) after FC and iSCSI? The play here is either for FCoCEE to replace standard FC, and coexist alongside iSCSI for the uses which go beyond a single subnet, or will iSCSI become the single storage protocol.

[Storagezilla]

I think the management tools thing stems back to expanding the functionality of existing FC SAN management tools, or something.

I don't see iSCSI going away in the same fashion I don't see the millions of SAN ports in production going away anytime soon either. The most likely scenario is that FC morphs into FCoE and over (A lot of) time as current fabrics age and are replaced we'll end up with a lot more FC traffic sitting on IP.

iSCSI does have traction in specific markets, EMC has pretty much iSCSI enabled all it's storage so ultimately it doesn't care that much which standard wins so long as EMC doesn't lose, but the fact that you're currently running your IP Network in parallel to your SAN is something a lot of SAN owners balk at.

This is one of those interesting technical debates which ultimately boils down to the question of what type of pipe you want delivering your water. Copper or plastic? Come to think of it does it matter so long as you get the water delivered at the right pressure, in the right volume, and at the right price?

Pure play vendors will always favor their flavor, but anyone else will just ship you the right connectivity hardware in the chassis and besides that you won’t see any real difference.