Can Storage Resource Management Save Us From Ourselves? We Quiz 9 SRM Vendors To Find Out.

Companies have an expensive storage hangover. Blame the way storage gear is sold, its proprietary nature, or IT's lackadaisical attitudes. Time to fix the problem--or go broke.

Jon Toigo

May 14, 2008

12 Min Read
Network Computing logo

As companies seek to make the most of IT investments while containing new spending, storage is coming under more scrutiny--and for good reason. Between 33 and 70 cents of every dollar spent on hardware goes to storage, and in most shops we're seeing allocation efficiency hovering at less than 20% of optimum for distributed systems storage. Utilization efficiency--using the right storage in terms of price and performance to host data based on its volatility, access frequency, business value, and other criteria--is in the single-digit percentiles.

And acquisition cost is only 20% of the total cost, according to Gartner. From the perspective of carbon footprint and infrastructure utility costs, storage is rapidly outpacing servers as the biggest user of power in the data center.

Could storage resource management, or SRM, be the key to halting this sprawl? The question has gained new urgency. When times were good and IT found itself short of space, there was a tendency to throw more capacity at the problem. Vendors were only too happy to reinforce this practice. Today, the financial times aren't so good for many industries, and power costs are racking up. It's time for IT to manage storage resources better.

InformationWeek Reports

Given that, we decided to do a sanity check on the state of the SRM software intended to help monitor storage system and interconnect status; automate routine configuration tasks; and, in some cases, facilitate movement of data across the storage infrastructure in accordance with policies.

We invited vendors to respond to a survey and asked what functions a management tool should provide and what criteria IT should use to evaluate SRM products now on the market. We also wanted to discover whether there's data that shows the value of storage resource management in cost savings or operational efficiency. In the early 1990s, analysts had argued that SRM would cut about 40% from storage cost of ownership. Others said a good SRM tool would let a single administrator manage 10 times more capacity than without storage resource management. Could vendors validate, or at least update, these claims?

DIG DEEPER

GET REAL

Thin provisioning, data deduplication, and compression are no substitutes for true storage management.

Download this InformationWeek Report

>> See all our Reports <<

We also asked what impact virtualization is having on storage--does it facilitate management or create more impediments? And what of the Storage Management Interface-Specification (SMI-S), the one-time darling of the Storage Networking Industry Association, which was touted as the path to storage management nirvana? Has it made any real impact on heterogeneous storage infrastructure management? If not, what hooks are SRM products using to monitor and interact with storage devices?

We received responses from BridgeHead Software, CA, Hewlett-Packard, IBM, Network Appliance, Northern, Symantec, and Tek-Tools Software, along with newcomer Olocity. And we sought observations from IT pros who've experienced the pointy end of SRM.

YOU'RE ON YOUR OWN
Storage resource management can't halt sprawl by itself. Data growth will, ultimately, drive storage capacity increases even in well-managed infrastructures. However, the reality is we can't see capacity clearly because of vendor "proprietariness" that's exacerbated by sleight-of-hand technologies, including thin provisioning, deduplication, and virtualization.

Because you can't manage what you can't see, SRM is supposed to give us clarity and visibility into what we have so we can administer it more effectively. That's one ingredient of a real strategy for containing sprawl. The other components fall under data management, including hierarchical storage management and granular intelligent archiving based on data class and policy. Combined, SRM and data management equal effective storage management.

The problem is, there is no standard set of SRM functionality (see story, "Key Functions That Define SRM"), so IT buyers need to have a clear set of objectives they're trying to address, lest they wind up with a bottomless project that devours time and resources like the storage infrastructure it was meant to tame.

"Identify where you plan to make savings that will pay for the product and deployment," says CA's Marco Coulter, VP of software engineering, and Todd Michaels, product manager.

Easy for them to say, but this goes to the heart of an issue long plaguing SRM. "The price tag for off-the-shelf SRM is too high, and so Perl and my backup server have become fast friends to give me some analysis reports on storage utilization at my company," says one IT pro, adding that he tried SRM from Veritas-Symantec but ripped it out in short order, preferring the more exact and less cumbersome reporting he can achieve using Perl scripts.

To address the cost issue, Coulter and Michaels emphasize the need for "reusability value" in the information collected by SRM tools, which provide the means to retrieve a lot of stats about storage and data. These metrics can provide a foundation for other tasks, such as compliance and governance policy making and monitoring, file classification, and storage tiering. While Tek-Tools CEO Ken Barth agrees with the reuse potential of SRM data, he takes the pragmatic position that ease of installation and the ability to solve immediate needs quickly are most important. Too many SRM purchases result in software gathering dust on the shelf because no one has the time, budget, or manpower to undertake a deployment.

Dean Snyder from HP's Business Server Automation Group articulates a list of criteria, including the number of agents required, depth of heterogeneous vendor support, integration with enterprise management tools, and business app and cluster/virtual machine awareness. Snyder also adds to the list "viability of the supplier," which strikes us as a marketing tactic intended to dissuade readers from choosing SRM products offered by startups and small companies, such as Olocity. Don't discount smaller vendors--many are thinking outside the proprietary box. One such interesting company is MonoSphere, which is set to release technology to tag storage with equipment value, so you can see the cost not only to store data on a platform, but of inefficient use of that capacity.

Pillars Of Storage Management
To get your storage sprawl under control, you need to own these four areas:

POLICY
MANAGEMENT

Map business rules to data and storage

Report on and forecast infrastructure trend

Impose process conformance with business as changes occur

DOES VIRTUALIZATION HELP OR HURT SRM?
Despite still-modest levels of VM adoption, a consistent trend in most vendor questionnaires involves the need to deploy SRM to support virtual machine environments from Microsoft, Virtual Iron, VMware, and others. In fact, Network Appliance's chief marketing officer, Jay Kid, puts "support for VM consolidation" on his list of top selection criteria.When we dug deeper into why claims about support for VMware, Hyper-V, or SAN virtualization figured so prominently in responses, the consensus was that virtualization neither helps nor hurts the efficacy of SRM; it all comes down to whether the virtual server or storage architecture was designed and rolled out with management in mind. Of course, storage vendors also are eager to take advantage of the increasing interest in virtualization by showing how their offerings can add value to virtual platforms.

"Competitive 2008 SRM tools will need to discover, report, and map storage resources that are attached to a virtual host," says HP's Snyder. Most vendors expect steadily increasing deployment of server virtualization, meaning better-than-even odds they'll run into storage attached to a number of physical hosts under virtual machine management. Lack of VM support, Snyder says, will make accurate capacity and utilization reporting impossible and produce topology maps that don't properly stitch VM hosts to their associated storage resources. This complicates provisioning, performance management, troubleshooting, and more.

The vendors are split over the impact of virtualization on SRM efficacy, with CA's Coulter arguing that storage virtualization, at least, is making storage management more complex and difficult without adding value. In fact, the return on investment expected from server virtualization may actually be compromised by the lack of storage resource management, according to Tek-Tools' Barth. "Studies have shown that the majority of companies that virtualize their servers are actually seeing an increase in the growth of their storage spending due to a lack of visibility into their storage environments," he says. Without SRM, virtualizing servers "is like driving with blindfolds on."

Sean Derrington, Symantec's director of storage and availability management, agrees, adding that virtualization complicates storage management by introducing a layer of abstraction that reduces visibility from the application to the spindle. It's incumbent on SRM vendors to map VMs to the hypervisor and provide application-to-spindle visibility. Virtual logical unit numbers also present management challenges that SRM products must address.

Jamie Clifton, BridgeHead Software's director of business development, takes a slightly offbeat view of the benefits of SRM in the virtual world: That virtualization helps companies keep legacy gear in play longer and, with the help of SRM, lets new applications use older devices and storage methodologies. SRM is a prerequisite, from his perspective, to obtaining real value from virtualization.

Northern Software CEO Thomas Vernersson agrees about the potential benefits accrued to virtualization and the additional layer of flexibility it provides within storage architecture design. Done properly, he argues, virtualization allows the storage infrastructure to be organized in a more logical way, in turn making SRM policies easier to design and deploy.

"HP no longer sees any major impediments for mapping heterogeneous storage infrastructure," says Snyder. "We're very satisfied with our progress."

Unfortunately, none of his colleagues who responded to our survey share his view that storage management nirvana has arrived, or will in the near future. The fact is, any vendor will promise soup-to-nuts storage management--if you buy only its brand of storage. For example, the tools HP uses for managing Compaq EVA storage are not portable to managing the storage it rebrands from others, such as Hitachi Data Systems. Similarly, EMC's ECC SRM tools work well with some of its product lines--DMX, for example--but not as well with Clariion. The idea of SRM is that it is an inclusive management approach, not an exclusive one. This is an area where standards from the SNIA's SMI-S effort were supposed to save us. However, while it has been adopted as an international standard, SMI-S hasn't delivered on unified storage management. The fault is not in the spec, says Olocity CEO Roger Reich, who worked on the management approach from its earliest days and now seeks to build a business around helping vendors and consumers deploy and use the technology. Rather, SMI-S is hobbled by implementation. "[SMI-S] interfaces need to be easier to deploy and need better quality and completeness," Reich says.

Without effective "providers" (the SMI-S component that's supposed to provide monitoring and configuration access to storage gear), the underlying infrastructure of SRM tools is a rat's nest of millions of lines of code that integrates another rat's nest of disparate vendor proprietary interfaces that have horrible reliability. Reich adds that conventional SRM interfaces have wildly differing degrees of function and usually render SRM software a net efficiency drag to an IT shop using multivendor configurations.

Reich's views are echoed by nearly all other respondents. John Foley, IBM's TotalStorage Productivity Center marketing manager, decries proprietary systems and the lack of APIs as a source of vendor lock-in--not that IBM is opening its APIs to the world.

The upshot: IT must insist that vendors provide in their SRM products streamlined mechanisms to collect information and interact with storage gear. Today, the best we can hope for is a kludge of "hooks" ranging from proprietary APIs to SNMP MIBs. Vendors remain reluctant to provide open management interfaces that might contribute to the commoditization of storage gear and facilitate replacement of Brand X's box with Brand Y's less expensive system.

SHOW US THE MONEY
Still, Tek Tools' Barth remains optimistic that server consolidation and virtualization technologies are providing an education to IT departments regarding the merits of centralized management. In his view, control is shifting to enterprise IT groups, which in the coming few years will compel storage vendors to make good on interoperability. BridgeHead's Clifton also finds reason for optimism in the increasing popularity of Web services and the move toward SOA.

But storage vendors have stonewalled in the face of other open standards pushes, so why should now be any different? It won't, unless enterprises assert themselves and demand standards. This is likely to happen only after IT is convinced that SRM will yield measurable improvements in the things that matter: operational efficiency, cost savings, and risk reduction.

For SRM to gain this credibility, organizations need statistics showing efficiency gains from enabling fewer storage admins to do more via automation. But none of the vendors we talked to could point to a definitive survey or empirical analysis supporting cost savings or efficiency improvements.

"I'm not aware of any analytical studies of the actual human efficiency gains of implementing and using a large SRM solution versus the cost of using individual hardware element managers and operating system features," says Olocity's Reich, with some frustration.

In the absence of meaningful data points, vendors resort to anecdotal evidence. We've all heard these pitches and have learned to take them with some skepticism. Northern Software's Vernersson cites an SRM study his company performed for one of Scandinavia's largest financial institutions. "It was estimated that 32% of primary storage capacity would be reclaimed during the first six weeks after deployment, and that growth rate reductions of more than 40% would push back initially scheduled capacity investments by more than 18 months. The first-year savings alone were estimated at more than $400,000."

A few vendors offer calculators--CA, HP, and IBM have ROI analysis tools, for example--to help customers predict their potential gains from SRM. Only Tek-Tools' Barth would go on the record to state that his Storage Profiler product delivers the 40% savings in operational costs touted by Gartner nearly a decade ago. He says his customers have individually and collectively realized at least that large of a reduction in total storage costs via a combination of operational savings and infrastructure allocation efficiency improvements following deployment of his product.

Jon Toigo is CEO of storage consultancy Toigo Partners International, founder and chairman of the Data Management Institute, and author of 13 books. Write to him at [email protected].

Continue to the sidebar:
Key Functions That Define SRM

SUBSCRIBE TO OUR NEWSLETTER
Stay informed! Sign up to get expert advice and insight delivered direct to your inbox
More Insights