(FC) protocol, and how it sees the brewing battles in the marketplace between Fibre Channel and iSCSI. Fibre Channel is getting a little long in the tooth these days and faces major competition from iSCSI once 10 Gigabit Ethernet (GigE) components are mainstream. Why is Fibre Channel still relevant and are there any new features or capabilities for Fibre Channel fabrics in development? Tom Hammond-Doel: Security is first. The Fibre Channel Security Protocol (FCSP) is going out there -- a good solid security mechanism in Fibre Channel.
Then there's Fibre Channnel-Base-T, which will allow users to run Fibre Channel over unshielded twisted pair cabling just like standard cable used with PCs today -- there should be products shipping for that within 18 months. It's significantly less expensive cabling, there's familiarity with it and ease of connection as well.
End-port ID virtualization -- this is a very, very cool feature. It makes anonymous identification to a server possible, which sounds kind of ominous, but what it's really enabling are blade server systems. It means you can remove and replace blades, and connect them back to the same storage without disruption, even though the end-port ID has changed. There are a few of these products already out there, but the next generation of products should be coming out in a year to 18 months also.
Then there's the fabric application interface specification (FAIS) -- an API [application program interface] that fabric switches will use to standardize the interface to the fabric, similar to SMI-S with storage arrays. The spec is complete -- we expect to see products handling and managing data through FAIS within 18 months also.
There's evidence that iSCSI is making headway, even in the enterprise storage market. How do Fibre Channel vendors see iSCSI? Is it a threat?Hammond-Doel: Definitely not. Headway is a relative term. If you look at growth figures from IDC and Dell'Oro [Group], by 2009 they're still projected to be only a small portion of the market. Fibre Channel, going well into the future, is still going to be the predominant storage network. Ten GigE first coming out is going to be fairly expensive still -- and if I can buy a Fibre Channel system that can give me the same performance on a separate network with better security, with the highest cost-performance ratio available, there's still going to be compelling argument for it.
But not having storage on a separate network is thought to make managing storage easier in some circles.Hammond-Doel: Running iSCSI over Ethernet might work in larger businesses where managers are trained and understand how Ethernet really works. But in the market iSCSI is trying to go for, departments and small companies, they may have administrators more trained in networking, or administrators more trained in storage, but not both. Where managers are more trained in storage, Fibre Channel might actually be more familiar. Ultimately, though, the basic protocol on which they run storage really isn't going to make much of a difference.
What will, then?Hammond-Doel: The ability to manage storage easier. Fibre Channel, from a performance standpoint, will always be the top dog so to speak. The market that is the great majority of where we have been is in the enterprise, but there has been an ongoing desire to make systems easier to implement. The capabilities of Fibre Channel will be brought into smaller offices or branch offices -- it is extremely important to us to make sure we reach those markets.
Several initiatives are under way, things like SCM -- simplified configuration management -- all focused on bringing Fibre Channel in a new direction, down into lower end markets. SCM isn't really a technological change -- it's more of a profile, a subset of the Fibre Channel protocol that makes certain devices simpler to use. A perfect example of a profile would be to specify that, even though the overall protocol specifies over 16 million possible nodes in an Fibre Channel network, SCM will say that only two switches should be on the A side or B side of the network infrastructure. The whole idea there is by maintaining a tightly focused set of requirements, it's going to be much easier for end users to be able to set up and get a Fibre Channel system running. That's one of the primary focuses of it, to assure the level of training for use of Fibre Channel SAN [storage area network] would be a non-IT manager type of person.
What about when 10 Gbit becomes more common? How will that affect the competition with iSCSI?Hammond-Doel: There's already 10 Gbps Fibre Channel today. Already out there and deployed -- the Fibre Channel protocol has two bases, base two and base 10. Base two starts off at 1 Gbps, then goes to 2 Gbps, then four, then eight and so on. Base 10 starts off at 10, and we're considering bringing it to 20 -- it'll go in multiple speeds of 10. The two bases have different underlying coding -- 64.66 with base two and AB10 with base 10. In both Fibre Channel and iSCSI cost and quantity of 10 Gbit makes it a minority product, and it's not expected to grow at any huge rate. Moreover, 10 GigE is being used mainly in interswitch and backbone communications -- where Fibre Channel 10 Gbps is used, it's going end to end, with switch-to-switch connections. In either case, the market isn't going to have 10 Gbps performance, either on Fibre Channel or Ethernet, going all the way to the desktop anytime soon, and the price is still high, regardless. In fact, it's right around the same price point, and both are effectively using the same electronics.
What kind of traction are you seeing in the market for 4 Gbps Fibre Channel? Has there been overwhelming adoption or is it a slow process?Hammond-Doel: The feedback I've been getting from companies is that the transition from 2 Gbps to 4 Gbps is distinctly happening faster than the transition from 1 Gbps to 2 Gbps. Once people got onboard with the first transition, it was a lot easier for them to make the move to 4 Gbps, even if they're still running some parts of their environment at 2 Gbps. It's all part of our future-proofing argument: We look at Fibre Channel as being the truly future proofed protocol. We are ensuring that we are forward and backward compatible in our speeds so each new generation of performance is compatible with two previous generations of equipment -- so 4Gbps is compatible with 2 Gbps and 1 Gbps and will be forward compatible with 8 Gbps and 16 Gbps when they come out. Any time anything new comes into the protocol, there's going to be an assurance that nothing is orphaned or unusable.
Why are they switching so readily? What's the benefit of 4 Gbps for them?Hammond-Doel: Customers are getting 4 Gbps at the same or even lower price than they were for 2 Gbps, so it is essentially free. Even if they don't need the full 4 Gbps right away, as customers migrate from 2 Gbps to 4 Gbps, they buy 4 Gbps to be ready as their deployments mature. More and more applications are demanding higher bandwidth, and as hard drives get larger and larger, the time it takes to rebuild a drive in the event of a failure gets dangerously long. Higher speeds help mitigate the rebuild time. Also, in general, society is moving from a PowerPoint to a Podcast mentality -- meaning a lot more audio and video content and a lot more management and movement of larger AV [audio/video] files.
What about the possibility of a Fibre Channel-SATA spec? There was an industry group lead by Emulex [Corp.] that was working on it. Are you working with them? Has there been any progress?Hammond-Doel: There's been considerable progress there. You might see SATA tunneling over Fibre Channel deployed in large enterprises sooner than SATA over SAS [serial attached SCSI]. Currently in the T11 organization, the process is progressing extremely well, pretty much within a month of original target dates. The August meeting is coming up next week, and the new spec will be going to letter ballot within the next three to four months. I'd say within a year you could see scalable SATA products over Fibre Channel.