In-Depth

A Storage Soap Opera

You’re supposed to dance with the one who brought you—it says here in small print.

• By Jon William Toigo
• 04/21/2005

Covering storage these days is like writing for Soap Opera Digest. In the past several weeks, we’ve had a flurry of announcements from Network Appliance, IBM, and Hitachi Data Systems that make very little sense to the outside observer. To be honest, the ramifications of these moves are equally unclear to people inside the industry.

Network Appliance and Hitachi Data Systems recently announced that they were breaking off their relationship in the network-attached storage space, which dated back to December 2002. In those happier times, HDS was working with NetApp to create a NAS head, dubbed gFiler, to front-end its Thunder and Lightning arrays. Apparently, the couple had some irreconcilable differences. In the announcement, HDS stated that it was abandoning that NAS initiative and instead was building a NAS blade that would install directly into TagmaStore, their “universal” storage platform. There was no comment on whether HDS would develop standalone NAS products going forward.

HDS's move might make sense if (1) TagmaStore were really picking up a head of steam with large enterprise consumers—something that doesn’t seem to be the case currently—and (2) if the company had new ideas for dominating the NAS world—which might well be the case—possibly keyed to an announcement made by its parent company a week or two before. Hitachi Ltd., the mother company, made headlines in late March by stating that it would deliver a 1 TB disk drive to market by 2007 that leverages perpendicular recording technology.

The simplest way to explain perpendicular recording is that it is a recording method in which bits are written on a disk so that their magnetic field is directed vertically—or perpendicular to the media—rather than horizontally. Such recording surmounts the problem of superparamagnetism that currently constrains the areal density—the number of bits you can squeeze into a square inch of media—of traditional disk to about 150Gb per square inch.

With traditional “horizontal” recording, as bits are placed closer together, you run the risk of their magnetic fields interfering with each other, with the result that bits “flip” at random from positive to negative states. This is a bad thing because it messes up the data that the bits collectively represent. By writing bits so that their magnetic fields emanate vertically from (or perpendicular to) the media, the bit fields don’t interfere with each other nearly as much, so you can squeeze a lot more data into the same space on the disk.

Hitachi says its perpendicular disk will be ready for market by 2007—just in time to surmount the limits to areal density growth that will be encountered in traditional disks. (Current disk densities hover around 40 Gb/square inch, but density is doubling about every 18 months in an inexorable drive toward the 150Gb/square inch superparamagnetic barrier). Given the theoretical capacities of perpendicular recording drives, which go well beyond 1 TB per drive, it is not beyond the imagination that a single drive, equipped with a journal file system and RAID 1/0 software, could deliver the same basic feature set more efficiently (and at a much lower price point) than any current generation Network Appliance Filer. If this kind of technology is on HDS’ drawing board, they certainly do not need their former buddies in NetApp land to make a big splash in the NAS market.

Faster than the ink could dry on a Britney Spears Las Vegas divorce, Network Appliance hit the newswires again to announce a new relationship—this time with IBM. As discussed in a phone briefing, IBM was phasing out its own NAS line and replacing it with Network Appliance Filers. Big Blue made no bones about the intention of the move: to crush arch-enemy EMC in the NAS market.

The weird thing is that EMC is certainly not the industry leader in NAS. More weirdness: NetApp followed this news the next day by announcing its acquisition of Alacritus, a vendor of virtual tape technology. This move was odd given that IBM has several virtual-tape products in its software inventory that might have been adapted by NetApp for its use.

On its face, the Alacritus acquisition seems to mimic the behavior pattern of a jilted spouse. Having been cut loose by HDS, NetApp might now be circumspect about investing its full trust in any partner. On the other hand, we may be reading too much into the Alacritus acquisition. For all we know, the acquisition may have been in the works long before NetApp’s parting from the HDS relationship and its partnering with IBM. Still, viewed as a sequence of announcements, the Alacritus deal seems to suggest that NetApp is cautious about going too far, too fast with IBM.

One can safely assume that, in any case, troubled waters lie ahead for the new couple. In the area of used equipment, for example, IBM has been golden for many years: extremely willing to support its own used gear with new licenses and warranties for customers who prefer to save a few bucks by buying used. NetApp, by contrast, is rivaled only by Sun Microsystems as the least-friendly secondary-market vendor on the planet. Buy a used NetApp box today and you often find yourself stuck with an array that cannot be attached to a network via NFS or CIFS. NetApp, as a rule, does not allow their software licenses to be transferred with their gear.

How will these philosophical polar opposites be reconciled? Nobody is willing to talk about it publicly, but it is clear that IBM has made a big business out of used equipment licensing and servicing, and, where they include NetApp in their customer solutions, Big Blue will likely expect their NAS supplier to fall into step. That would be a good thing for consumers.

The final announcement in this flurry had IBM’s name in the title. In early April, Big Blue announced that it had shipped 1000 virtualization solutions. More to the point, they had shipped 1000 more virtualization solutions than had EMC. Despite their PR department’s cajoling, we found no real story to report in the IBM announcement, since EMC’s promised Storage Virtualization Router hasn’t shipped from Hopkinton yet.

What might be more newsworthy are the reports we are receiving from consumers who are complaining that their IBM SAN Volume Controllers are introducing performance problems in some installations. Most of the complaints we are hearing have to do with the implementation of the technology on feature cards deployed in Cisco Systems Fibre Channel fabric switches. If you are a user of the SVC and can update us on your experience—positive or negative—with the technology, we welcome the input at jtoigo@toigopartners.com.