iSCSI's Second Act

The storage networking industry's prodigal son, iSCSI, is back. Now that the Internet Engineering Task Force (IETF) has officially ratified the iSCSI specification, high hopes for IP SANs have returned faster than a fatted calf on methamphetamines (see iSCSI Gets Go-Ahead and IP SANs: Coming of Age).

But this time around, IP SAN vendors have set the bar several notches lower. Over the past year or two, they have eaten humble pie, gotten the wakeup call, had their wings clipped pretty much all at the same time. No longer do they say iSCSI is a "disruptive technology" that will uproot Fibre Channel and send it caterwauling into the chasm of obsolescence [ed. note: cue the Philip Glass score]. At least, not yet... more on this in a second.

In just one example of this retrenchment, Cisco Systems Inc. (Nasdaq: CSCO) earlier this year drastically cut back its iSCSI router division. That group was formed out of NuSpeed Internet Systems, a startup Cisco acquired for $450 million less than three years ago. Talk about a letdown! (See Cisco Lays Off NuSpeed Team and NuSpeed Duo Departs Cisco.)

Why has iSCSI been such a disappointment? The usual suspects range from stagnant IT spending in general to a longer-than-expected standardization process for iSCSI (see iSCSI in Exile).

What it boils down to, though, is that iSCSI over Gigabit Ethernet does not offer any overwhelming advantages over 2-Gbit/s Fibre Channel – except cost. Yeah, that's a biggie. Maybe even a huge-ie in the grand scheme of things. But is that enough to be disruptive? Not yet."Only where connectivity and low cost of installation are of paramount importance will iSCSI succeed at 1 Gbit/s," write Gartner Inc. analysts Nick Allen and James Opfer in a research note published last fall.

Today, iSCSI simply doesn't offer enough to induce people to tear the walls down and pull out the pipes, says Bruce Hillsberg, director of storage software strategy in IBM Corp.'s (NYSE: IBM) systems group. "The iSCSI-vs-Fibre Channel decision is like whether to use copper piping or PVC piping," he says. "It's really a question of which technology better suits your needs. But unless you're building a new house, you're going to keep the plumbing you have."

We're now coming out of the low point of the trough in the iSCSI hype cycle. Instead of conquering the SAN market in one fell swoop, iSCSI's Second Wave is about two things:

So what does this mean for how the IP SAN market will develop?

The first area is interesting, but it doesn't strike me as one that will remain discrete for very long. Why? Because it's a stopgap measure, not a long-term, strategic architecture. You can drop in one of Cisco's iSCSI-to-Fibre Channel routers to connect "stranded" servers via Ethernet to an existing SAN. But you wouldn't design a SAN like that from scratch if you didn't have to. And anyway, all of the Fibre Channel switch players expect to provide iSCSI or FCIP (or both) into their platforms, making this more a feature than a product (see Sprint Puts Cisco to Test, Brocade Reupholsters Rhapsody, McData's Psyched About Software, McData Demos FCIP).The second area – native IP SANs – represents a bigger long-term opportunity. But it will start small. Intel Corp. (Nasdaq: INTC) understands this: It sees the arc of iSCSI starting with low-end storage systems designed for small and medium-sized businesses that can't afford a Fibre Channel SAN, either in terms of buying it or managing it. Intel views the world in the PC model: As volumes increase, iSCSI-based systems will become cheaper and more robust, scaling up into higher-end markets (see our interview with Mike Wall, GM of Intel's Storage Components Division).

The other issue is that IP SAN vendors aren't selling iSCSI. They're selling SANs.

"Customers don't wake up and say, 'This is the day for iSCSI,' " says Peter Hayden, CEO of EqualLogic Inc., which is developing an iSCSI-based storage system. "Customers wake up and say, 'I've got to build a SAN.' " The challenge is conveying the benefits of storage area networks to an entirely new class of customers.

So far, it's been startups like EqualLogic, Intransa Inc., Sanrad, and StoneFly Networks Inc. that have been the most aggressive at marketing iSCSI-based systems. But more established players, including Intel, Network Appliance Inc. (Nasdaq: NTAP), and Adaptec Inc. (Nasdaq: ADPT) are starting to get in on the action as things start to coalesce. (See EqualLogic Tallies $15M, Intransa Cranks Up IP SAN, StoneFly Offers 'Complete' IP SAN, Intel & Emulex: iSCSI Tag Team?, NetApp Blitzes on iSCSI, and Adaptec Details IP SAN Betas.)

The only question left on iSCSI is timing (although perhaps that's been the only question all along). IBM had its own false start with the iSCSI-based 200i storage device, which it gently phased out last year once it became clear no one was interested in buying it. Big Blue's current strategy is to support iSCSI across its entire range of storage products, but it will "enable that in the timeframe with the rest of the market," says Hillsberg (see IBM Ditches iSCSI Box).I think the deep-freeze surrounding iSCSI will finally start to thaw in a noticeable way in June, once Microsoft Corp. (Nasdaq: MSFT) releases the final version of its iSCSI driver for Windows (see Microsoft to Unleash iSCSI).

And remember, this is just the start of Act II. There will be an Act III in the iSCSI drama – that will happen once 10-Gbit/s Ethernet becomes widely available. At that point, while Fibre Channel will still be an actor on the SAN stage, it may find itself in a steadily diminishing role.

— Todd Spangler, US Editor, Byte and Switch