In-Depth

NuView Sequels Keep Coming

StorageX answers a specific problem with a good solution. But how can the company cope with a niche solution that refuses to remain just a niche solution?

• By Jon William Toigo
• 08/28/2003

The secret to his success seems to be his tendency to focus on a specific end-user requirement, developing a non-disruptive, best-of-breed solution, then working out OEM deals with “brand name players” that need his technology. If you have ever used SQLAssist (acquired by Platinum, then part of Computer Associates offerings), ManageX (acquired by Hewlett Packard Company), ClusterX (acquired by Veritas Software), or the company’s current StorageX offering (OEMed by Network Appliance as VFM), you know Mehta’s work.

After chatting with Mehta and his new Director of Marketing & Strategy, Apurva Patel, recently, I think this NuView sequel also has big potential. Simply put, StorageX answers a specific problem with a good solution. Mehta’s only problem appears to be how he will cope with a niche solution that refuses to remain just a niche solution. Let me explain.

In creating StorageX, the NuView folks set out to address the problem of storage scaling in the Windows file system world. Specifically, they wanted to find a way that would enable a consumer to scale file system-based storage without the hassles of either (a) keeping track of a bunch of directories scattered over a bunch of boxes, or (b) figuring out some expensive and proprietary mechanism for growing storage volumes so that more files could be stored on the same volume.

We aren’t talking about block-level database-focused storage scaling here; just files. If the folks at UC Berkeley are correct, most of our storage growth today is occurring in end-user file systems, and not in databases (where so much work has been done to address scaling issues with SANs). Typically, the increasing storage demands of file systems require the deployment of more servers and direct-attached arrays, or more and more network-attached storage appliances. With each new deployment, files and directories proliferate. The job of administering which users see which directories and which files is a nightmarish one. The pain is so great, according to Mehta, there is a growing business in “global namespace design” consulting—that is, designing a strategic map of directories to meet company information access requirements.

To deal with this problem, Mehta considered many alternatives. He decided that growing bigger volumes through hardware- or software-based device virtualization was only a stopgap, and an expensive one at that! On the other hand, attacking the problem with a replacement file system—a distributed FS that could mask complexity across many boxes—was also a flawed strategy. To garner favor with the consumer, you needed to scale horizontally, but also non-disruptively.

Mehta’s solution is StorageX. It is essentially a metadata server that sits outside of the data path between servers and storage devices where it provides a home for a unique global namespace. Operating like a directory server, it presents intelligent file directories to end users and applications and brokers their access to the cluttered back end storage platforms and the tangle of volume names, shares, and directories that exist in the messy storage infrastructure.

You don’t replace your file system on the storage device or on the server in Mehta’s approach. It is non-disruptive. The StorageX server simply acts like a directory server in an n-tier client/server application.

The problem with this technology, however, is its refusal to remain a niche solution. Network Appliance OEMs it as “Virtual File Manager” because it lets you simplify the presentation of files stored on a bunch of NetApp Filers and provides an intuitive, readily managed, scaling solution that doesn’t impact its cool-but-proprietary Write Anywhere File Layout (WAFL) file system. But StorageX keeps finding new ways to exploit its file management paradigm.

Mehta and company are now adding things like byte-level differential replication for improved data protection. StorageX can replicate data heterogeneously and across multiple locations, and send only byte-level changes over the WAN to save bandwidth. When a storage device fails (or an entire facility experiences a disaster), you can simply fail over all file pointers to an alternate set of storage devices in a different location.

Mehta is adding data lifecycle automation to the solution as well, enabling it with a policy engine that can migrate files and directories around based on last date modified and other criteria. Pretty neat stuff.

If Microsoft delivers on its plans to replace its file system with a SQL database in Longhorn, StorageX may once again become a niche—used primarily with legacy file storage. But, until that happens, StorageX is worth a look today—especially, if you have any pain in file system management. Who knows: by the time we convert files to objects, Mehta will probably have added new bells and whistles to his product to address the pain in that environment as well!