News

Opinion: Microsoft vs. Network Appliance: Full of Sound and Fury, Signifying Nothing

• By Jon William Toigo
• 06/26/2003

While I am hesitant to count out NetApp as one of the leaders of the $5-plus billion NAS market, especially at the enterprise level, I have to marvel at the obvious "marketecture" surrounding the story. The Microsoft play demonstrates clearly that network-attached storage is a misnomer, that NAS is still server-attached storage: it is just a server OS bolted to a storage array to give the appearance of one seamless unit, rather than a separate server and array. Practically speaking, there is little difference between the two configurations.

The Microsoft offering doubtless capitalizes on the improved storage management tools in Redmond’s upgraded OS. However, Network Appliance has also been improving its management capabilities and its integration with Microsoft OS and file system requirements. I fail to see a compelling argument to support the claim that Microsoft environments are better served by one platform or the other.

While it may give comfort and solace to IT managers in “pure Microsoft” environments that even their NAS platforms have “Microsoft inside,” why is this a major benefit? Given that Redmond has already stated that it is abandoning file systems with its next generation OS platform, code-named “Longhorn,” due out in another 12 to 18 months, it seems to me that buying the latest platform from Redmond is like deploying a Fibre Channel SAN, which is already referred to as a legacy SAN in the face of burgeoning IP storage networking technology.

From the perspective of NAS evolution, the new Microsoft offering is not a bang but a whimper: an incremental improvement in a technology that always seems to be undergoing incremental improvements. I just hope that Redmond does not take to emulating the silly product positioning of other vendors in the storage industry and oversell its latest generation product. When Longhorn is released, the company will have something real to crow about. For now, it would be a bad idea to spend a lot of cycles or a lot of money differentiating itself from NetApp or other established players.

Longhorn changes the rules of the NAS game. NAS will no longer be part of an industry-initiated storage bifurcation scheme that separates files and blocks into two separate categories of storage requiring two different kinds of storage. Longhorn substitutes an object database for a file system. Everything becomes blocks again. Only then does legacy terminology such as “filer” go by the wayside.

In the meantime, performance is what you should care about. If you are seeing only incremental improvements in NAS architecture, then focus on how well the product does its job—the size of its memory cache; the speed of its back end connections to its storage device trays; and additional functionality the vendor has added for management, backup, and encryption of data. These are important features and functions that nobody is talking about.

Another thing you should care about is price. At a time when we are all more price sensitive, a meaningful evolution in contemporary NAS would be one that contributes to a lower solution cost, both in terms of acquisition and product cost of ownership and lifecycle.

NAS is increasingly a commodity and vendors should drop the entry cost for NAS to more accurately reflect the pricing of its commodity components. I’m all for rewarding secret sauce intellectual property, but I want to see the preliminary price of NAS products reflect a bare bones configuration that I can then enhance with whatever bells and whistles I think I need.

Moreover, I want somebody to explain to me why I want to pay a high price for a car that has no transferable warranty and loses 50 percent of its value as soon as I drive it off the lot. This metaphor applies to NAS in spades. Like most automobiles, most NAS boxes are built from similar parts. Most NAS products use cut-down UNIX or Linux kernels bolted to either parallel SCSI-, Fibre Channel arbitrated loop-, or FC fabric-connected arrays of disks. They connect to open or quasi-open standard network protocols such as NFS, CIFS or HTTP. So why is it that the NAS software for making these back-end and front-end connections can’t be transferred, together with the hardware, to somebody who wants to buy your older gear on, say, e-Bay?

The bottom line is this: the manufactured contest between Microsoft and Network Appliance in the trade press is largely irrelevant both from the grander perspective of NAS evolution and from the practical perspective of consumer requirements. Don’t let yourself be sucked in.