Microsoft imperialism could benefit storage

Storage virtualization has long been the Holy Grail of many network administrators, giving them the ability to simply see all storage capability on the network as logical-layer volumes without regard for where the storage device is located, how many discrete disks it represents, or even what media it might be.

This not only represents a much easier user interface to network resources, it nicely divides administrative skill sets such as backup and storage management with that of physical layer systems management. The problem is that vendors have been unable to agree on what constitutes a universal driver — that magical piece of low-layer software that will run on any OS platform, and cleanly mark the divide between disk and DOS. Third-party software manufacturers such as Veritas or even IBM’s Tivoli have attempted to create a software management layer between disk and OS to simulate storage virtualization, but none of these have ever completely pulled off the full virtualization vision.

Anytime standards are an issue, we’ve come to expect Microsoft to bite like a shark attracted to wounded prey. But until Windows 2000 Server, even Redmond didn’t want to get involved in the complexities of storage virtualization. Since then, however, it has embarked on an iterative quest to bring this kind of disk-neutral storage virtualization to the Windows platform, and Windows 2003 Virtual Disk Service is the latest incarnation.

As you might expect, all Microsoft has done here is to provide a VDS (virtual disk service) API that should allow third-party storage vendors to create products that Windows 2003 Server can treat as virtual storage volumes. I say “should” because thus far such products still haven’t seen the light of day, so we can’t really be sure how well the VDS API actually works.

And even if it does work, it’s a Windows-only brand of storage virtualization, so that must be bad, right? If you’re a storage purist, I suppose you can take that stance. As a long-time purveyor of Windows network administration, however, I’m slightly pink-cheeked when I say that it really doesn’t bother me so much. If the VDS API will allow me to offer my customers virtualized storage, which a surprising number of them have been asking for, then I’m all for it.

Even though most are still running heterogeneous networks, they’re still usually about 90 percent Windows-based. If I can virtualize storage on 90 percent of a 500-node network, that’s still a big step forward. And if it spurs the Linux community or IBM to develop similar products for their operating layers, that only makes my life easier, even if I still have to segregate those resources on a large network. Three storage segments are still easier to manage than 10 or 20 departmental ones, for example.

Where Microsoft and its third-party providers will need to do some work is in competing with what already exists. My 90 percent reasoning isn’t new; Veritas, Tivoli, and others thought that up a long time ago and created software layers to manage this very process, most aimed heavily or even solely at the Windows platform. Where they had trouble was in pricing and additional complexity for administrators. If Microsoft’s storage partners want to compete in this space with a “native” solution, they’d better be prepared for a war, and their weapons had better be seamless integration, true ease of use, and a low-blow price tag.