Storage Pipeline: Review: Out of the Abyss: Page 4 of 8

The three vendors also rallied around the battle cry of automated
provisioning. There were subtle differences, however. For Fujitsu Softek, whose
solution set includes an in-band virtualization engine (a customized version of
DataCore Software's SANsymphony product), provisioning seemed to mean creating a
pool of unallocated physical storage capacity from all managed arrays, whether
connected in direct-attached storage, network-attached storage or SAN
topologies, and then parsing it out as virtual volumes that could be scaled on
command behind applications.

Storability also suggested this approach. Its proposal referenced "an
optional Catalyst module" that might be used in connection with its Global
Storage Manager SRM product to create scalable virtual volumes via array "LUN
masking." Storability would also support certain third-party LUN (logical unit
number) masking products or virtual volume managers, like Veritas Volume
Manager, if they were already in use at Minuteman.

Declarations about the necessity of storage virtualization for effective
provisioning were notably absent from the Computer Associates proposal, however.
In response to our question about how provisioning would be automated, CA
instead discussed data migration. The vendor seemed to be treading a thin line
between conventional wisdom about the need for virtualization and what its own
experience had taught it about the amount of storage lost to poor management in
general. Although CA refrained from making explicit statements, its response
implied that provisioning in most shops wouldn't be nearly the pain that it has
become if a system of policy-based data migration were applied rigorously and
capacity managed properly in the first place.

One might interpret CA's response as an effort to deflect attention from the
fact that the vendor has no virtualization solution of its own. However, it's
reasonable to assume that CA's BrightStor will support most any third-party
virtualization approach, so the decision not to mention virtualization in the
proposal seems deliberate. It is more likely a reflection of the vendor's belief
that disciplined and informed data management--including policies and procedures
for culling junk and stale data from primary storage--is a more effective route
to capacity allocation efficiency than is the addition of complex virtualization
technology.

Although Minuteman's problems might be addressed in the short term by a
dollop of virtualization technology, such a strategy might not solve its
capacity conundrum in the long run. In the final analysis, we preferred a
strategy focused on policy-based provisioning without virtualization.