Here's a hint: It's not Cisco, IBM, Intel, or any of the dozens of start-ups.
By John Webster
You may not think iSCSI needs to be saved, what with the market's renewed attention on the IP-based storage I/O alternative to parallel SCSI, Fibre Channel, and Serial ATA. Yet we've been here before, haven't we?
Remember three years ago when iSCSI was the next new, new thing? It was going to be cheap. It was going to be ubiquitous. After all, IP was everywhere. iSCSI would go everywhere Fibre Channel hadn't gone, and then it would have Fibre Channel for dessert.
IBM was pushing the protocol with products, partnerships, and services. And Cisco was throwing its considerable weight behind iSCSI, too, leaving many with the impression that Fibre Channel was a mere trifle to the networking giant and that its laser-guided focus would remain steadfastly on what it knew best. Cisco would enter the storage world through the door marked IP. So loud was the hype that at least one prominent financial analyst was pronouncing Fibre Channel dead.
Why, then, are we now—three years later—looking at what amounts to a re-launch of iSCSI? What happened? What brought the first iSCSI juggernaut to a grinding halt?
First and foremost, Fibre Channel turned out to be very resilient in spite of interoperability and cost issues. And then there was the often-overlooked fact that storage buyers are largely storage-centric, not networking-centric. We predict that if you ask a group of storage administrators if they think "storage networking" is a "storage" or a "networking" technology, a clear majority would say "storage." To them, iSCSI hailed from the networking side and thus had some proving to do if it wanted a place in the storage world.
Curious users (commonly colleges and universities) put iSCSI into test environments, but few users within commercial IT settings were even willing to take a flyer on the technology.
Those who did rarely moved iSCSI beyond the test phase and into production. And so the early ground swell of interest flattened to a barely perceptible ripple.
Then there was the IPSec speed bump. Just as Version 1 of the iSCSI specification was being spiffed up for its debut, the Internet Engineering Task Force (IETF) mandated that vendors somehow bake IPSec into the specification.
Shock and "arrrgh!" ensued—shock because the move was clearly not anticipated by many, and "arrrgh!" because vendors hadn't a clue about how to comply with the IETF mandate. At least one wheel had fallen off the iSCSI race car.
So, while proponents took a deep breath and regrouped, iSCSI languished. IBM pulled an iSCSI disk array off the market—a move that the trade press (erroneously) saw as the white flag of surrender.
Cisco then did its version of shock-and-awe with its monumental MDS 9000 switch platform/VSAN announcement, which was all wrapped around Fibre Channel. iSCSI was reduced to a mere footnote.
Now, iSCSI needs a new prime mover to put momentum back on course.
Don't look for IBM to retake the vanguard position. Once bitten, twice shy. Cisco is in no mood for distractions either, now that they are Fibre Channeling their way into the hearts and minds of storage buyers.
Who, then, to turn to? Could it be...would you believe...Microsoft?
It is true. Microsoft is now leading the iSCSI charge. It is populating its world with iSCSI drivers baked into Windows 2003 Server. Legions of Microsoft users now get iSCSI drivers for free. Microsoft has even put itself into the unusual position of being the only organization to publicly certify product compliance with the iSCSI specification. No one else, not the Storage Networking Industry Association or even the iSCSI Consortium, has publicly certified iSCSI interoperability.
Breaking down the barriers
Microsoft is breaking down the barriers to iSCSI adoption. Why? Because iSCSI solves problems for Microsoft, particularly those related to Exchange. Microsoft knows that direct-attached storage inhibits the growth of Exchange databases, and it has never been comfortable with the network-attached storage (NAS) alternative despite Network Appliance's success in consolidating Exchange databases onto its NAS filers. And Fibre Channel SANs have not been an attractive alternative for Microsoft either because of the cost and seeming complexity of implementing them. Microsoft likes mass-market opportunities, and Fibre Channel SANs are not mass-market plays.
On the other hand, iSCSI SANs are block-access-oriented ("a must" for Exchange, Microsoft says) and they're scalable. Once again, iSCSI SANs can be cheap and ubiquitous (read: mass-market opportunity here) because Microsoft can make it so. In fact, we believe Microsoft can single-handedly re-create the big iSCSI boom that eluded Cisco and IBM. And this time around, Microsoft might even get Intel to join in.
John Webster is senior analyst and founder of the Data Mobility Group (www.datamobilitygroup.com) in Nashua, NH.