Dive Into Pooled Storage

Systems administrators with experience working in mainframe environments have good reason to demand storage management automation. After all, they're used to it. The trouble is that most of them now work primarily, if not solely, in distributed computing environments. As Scott Kennedy, VP of strategic business development for storage management software vendor Fujitsu Softek (Sunnyvale, CA) puts it, "Many of the problems with storage management in the open systems arena were solved ten years ago in the mainframe arena." Computer Associates' Senior VP for Storage Management Nigel Turner agrees, noting vast differences in storage utilization, a key indicator of storage management effectiveness: "In centralized, mainframe environments, utilization is often 80% or 90%, while in the distributed space, it typically remains at only 30% or 40%. The difference is that, in the mainframe space, there are tools that monitor utilization in real-time and react to changing capacity demands automatically."

Fortunately, storage management software developers targeting distributed computing environments are knocking down the primary barrier to automation - vendor heterogeneity. "In a typical distributed infrastructure, there will be four or five operating systems, with multiple applications running across those platforms," says Kennedy. "When you add in software tools from each of the storage device vendors, you're looking at as many as 50 or 60 software products an end user would need to be trained to use."

So, storage management software vendors are building layers on top of the device-management tools provided by hardware vendors. Except for occasional tweaking at the firmware or microcode level, an overarching management layer protects users from having to deliberately access specific vendors' tools. That gives administrators a single interface from which they can set and monitor policies that automatically control a range of storage management functions: monitoring capacities, provisioning storage for applications, mirroring volumes, archiving files, and coordinating backup processes. "Automated storage tools should be designed to relieve users of the burden of dealing with individual devices," says Erez Ofer, EVP, open software operations, EMC Corp. (Hopkinton, MA). "Instead, users should be moving more and more into the mode of managing storage pools."

Industry Standards Ease Device Management
Of course, the job of "managing the managers" can be accomplished only if the managers agree to be managed. In other words, unless device vendors support attempts at defining interoperability standards, an overarching software suite has only limited ability to "take control" of underlying device tools. Not surprisingly, you'd be hard-pressed to find a vendor rep who doesn't confirm the company's commitment to and participation in standards initiatives. And, just take a look at the roster for any of the various working committees sponsored by the Storage Networking Industry Association (SNIA). They're packed with engineers and executives from the major hardware and software manufacturers.

Those committee activities reflect general movement on the interoperability front. As EMC's Ofer observes, "Vendors are exchanging APIs [application programming interfaces]. And, CIM [common information model] demonstrations have been occurring for a couple of years." Computer Associates' Turner admits that, when vendors write to standards, storage management software tools can readily interoperate with storage hardware. "We've created a standard CIM portal," he explains. "When our software encounters a device that has a CIM interface, it can immediately launch that device's management tools. So, building to CIM or Bluefin standards is a necessity. People are spending an inordinate amount of time getting information from various management tools."

But, even Turner knows that interoperability standards won't completely eliminate the need for systems administrators to deal with proprietary vendor tools - not soon and probably not ever. Why not? Well, business is business, after all. "In addition to building CIM interfaces, every hardware manufacturer will continue to also build proprietary extensions into their products in order to differentiate themselves from the competition," Turner says. "So, we'll probably never get completely away from low-level coding."

Although he doesn't point to the market as the primary factor, Fujitsu Softek's Kennedy agrees that systems administrators will likely always have occasional needs to access vendor-specific management tools. "When you buy, say, a fabric switch, you also get a specific tool to look into it. You're always going to need those interfaces to go in and upgrade the microcode," he says. "Higher-level storage management software just tries to simplify the view by embedding the path to the device. Since 90% of the time you're not going to need to run the vendor's software, making it accessible through a single management interface keeps you from having to find the device or remember how to launch its tools."