Storage Management Quarrels
The many levels of storage
In a truly creepy scene from the 1991 motion picture, "The Silence of the Lambs," newbie FBI Agent Clarice Starling visits Dr. Hannibal Lecter, a convicted serial killer so dangerous that he is incarcerated behind glass walls rather than prison bars in the basement of a hospital for the criminally insane. In the scene, Agent Starling offers Lecter a survey document used to collect information on the psychology of serial killers for the FBI's behavioral science database.
With Anthony Hopkin's famous bone-chilling stare, and a silky voice filled with irony, Lecter mimicks Starling's slight West Virginia drawl when he responds, "Agent Starling, [do] you think you can dissect me with this blunt little tool?"
The scene captures on many levels the problem with storage management and hints at why no truly standards-based approach for managing storage has yet become prevalent. Storage vendors will likely bristle at the comparison with a criminally insane antihero, and I'm not suggesting that they actually eat their customers. However, each time a standards-based approach to storage management has appeared in the marketplace—two examples of recent management standards that come readily to mind are the Simple Network Management Protocol (SNMP) from the Internet Engineering Task Force and the Common Information Model (CIM) under development at the Storage Networking Industry Association—storage vendors have been quick to add their verbal support for the efforts but slow to change their platforms.
The reason is simple. If a common management standard capable of managing all storage platforms were to take hold, vendors fear that it could contribute to the commoditization of all storage products. As a rule, storage vendors jealously guard even the slightest differentiator that enables them to discriminate their platform from their competitors' offerings. It just wouldn't fly to say that a half-million-dollar EMC Symmetrix array or an eighty-thousand-dollar Joe's JBOD (Just a Bunch of Disks) could be managed equally well by a common SNMP management information base (MIB) or CIM object.
Despite the well-intentioned efforts of some members of the Storage Network Industry Association (SNIA) to advance CIM-based storage management, that goal won't be realized without some cataclysmic reordering of the universe. And that isn't likely to happen until storage customers make it happen by demanding management standards compliance from their preferred storage vendors—something they haven't done yet.
Where Things Stand Today
In place of standards-based storage management, we have a balkanized environment made up of several layers of management software technology. At the lowest level of this software caste system are the point products, so-called because they do one thing and one thing only, usually in only one server operating system or storage platform environment.
Point products are not to be sold short. Truly excellent software products exist for defragmenting a hard disk, creating a disk partition, establishing a RAID level, or performing any number of other discrete tasks that are the "liver and fava beans" of day-to-day storage management. Most storage managers select from among the best of these products to create a tool kit for managing their storage.
The problems with this "quiver of arrows" approach, however, are threefold. For one thing, each tool has its own learning curve. It takes time to sustain and maintain proficiency with each tool used as the software evolves over time.
Second, the use of multiple tools from multiple vendors obviates the ability to automate management tasks, so point solutions rarely provide a means to enable more storage to be managed by fewer personnel.
Finally, there is the problem of interoperability. In the realm of SANs, we've already seen that using different products from different vendors to, on one hand, create virtual volumes from an assemblage of physical disk drives and array partitions, and on the other hand, to back up and restore files from and to those virtual volumes, can lead to serious problems if the two point products don't work and play well together. The result can be a 1TB file restoral process that requires over 100 hours to complete. All the Chianti in the world, to paraphrase Hannibal Lecter, can't numb the pain produced by such a predicament.
Seven Tips on Storage Management Software
- Be aware of the limitations of storage management software products before you purchase.
- Deploy management products in phases and test interoperability carefully.
- Seek products that support standards, even if your hardware isn't instrumented for standards-based management currently. Some products are, and yours may be at some future point.
- Vote with your checkbook. Insist on product compliance with standards. Remember that vendors have hit hard times recently too, and it's a buyer's market today.
- Don't fear frameworks, but check customer references to see what you can expect from product deployment.
- Look for hidden costs: Licensing fees aren't the bottom line. Find out what implementation services, training and agent add-ons will cost.
- Look for desirable functionality beyond basic monitoring and configuration. BMC does a great job of providing views of storage resources organized by the application or business process that uses the resources. That kind of "application-centric" or business process-oriented view can be worth its weight in gold.
SRM Products Proliferate
The current trend to address this problem is to gather a multiplicity of point products into an "integrated suite" for storage resource management (SRM). As network management expert Jill Huntington-Lee—lately of Tek-Tools in Houston, Texas—is fond of pointing out, integration is a marketing term that lacks tangible meaning. Most SRM products in the market combine a few best-of-breed point products with several also-rans. The resulting products often feature little more than brochure-level integration.
The SRM software vendor community won't agree with this, of course. One very prominent vendor in the storage resource management software space has been on a buying spree over the last couple of years, gobbling up other point product software vendors and their products. While the company's president concedes that the products are at least two years away from being fully integrated, that hasn't stopped the vendor's marketing machine from selling the mixture of products, many of which are incompatible on certain levels with each other, as the industry's first "seamlessly integrated" storage resource management software solution.
Customers and integrators have gone public with their complaints about the software's inadequacy and the inaccuracy of vendor marketing claims, but the vendor's market share continues to increase. Hannibal Lecter might be amused by the irony.
Frameworks: Problems and Prospects
Above the SRM products in this caste system are the storage management frameworks. Frameworks are essentially agent-based status monitoring tools. Some provide the capability to remotely operate the configuration tools of the managed hardware platform. Still others provide a "launch pad" for point products and SRM tools. The latest entries in this category are EMC's Control Center/Open Edition with WideSky (see January's column, "A Kinder, Gentler EMC?") and Tek-Tools Storage Profiler, which, while self-described as an SRM solution, is actually evolving toward a framework package.
Frameworks, in general, endeavor to do the same thing as standards-based management approaches such as SNMP and CIM—albeit in a standards-less world. They implement a common (but proprietary) agent approach to monitor storage platforms, then normalize the information collected by agents for display on a centralized or Web-based management console.
Less sophisticated packages stop there—with simple monitoring. Some vendors endeavor to increase the functionality of their packages by gaining access to internal configuration controls of the monitored hardware so that corrective actions to some identified troubles can be tried without operator intervention. Framework vendors say that automating certain tasks, and administering the performance of certain point product functions performed on a routine basis, can reduce storage cost of ownership and enable more storage to be managed by fewer hands.
Truth be told, few framework vendors actually call their products "frameworks." The reason for this goes to some of the deficits with the technologies in the past. Only a scant three years ago, the trade press was filled with stories voicing concerns about "fat agents" and "management agent proliferation." The former referred to the fact that some agents consumed too many cycles on a server, reducing the performance of the platform they were fielded to monitor and protect—rather like most desktop anti-virus tools today. The latter, agent proliferation, became an issue because of fears that management agents would clog the arteries—the networks—that interconnected the vital organs of the enterprise IT infrastructure. These concerns, while largely irrelevant today in the wake of faster processors and bigger network pipes, cast a pall on framework-based management solutions.
Framework also became a dirty word because of the difficulties many customers endured when rolling out their framework products. Setting up agents for managing heterogeneous equipment environments remains a cumbersome and time-consuming undertaking with many products today. With the exception perhaps of BMC Software's PATROL, high-end management frameworks can take months to deploy, even with expensive consulting assistance.
Worse yet, customers often perceive their carefully constructed management framework system as vulnerable to vendor change. Not only must software be updated as framework vendors make annual revisions to code, but there is always a possibility that hardware supported by the framework product today will not be supported tomorrow.
The reason is simple: To deliver their management capability, framework vendors must develop working relationships with equipment manufacturers that provide them with access to the command line interface (CLI), configuration tool application programming interface (API), or other mechanisms provided on the equipment for managing the equipment itself. These working relationships can become strained and can be broken if the framework vendor and equipment vendor have a falling out.
This actually happened in the late 1990s in the case of Cabletron Systems (the vendor of a network management system product called Spectrum) and Cisco Systems. Cabletron, which also manufactured intelligent network hubs, adopted a marketing posture that criticized Cisco routers and switches. Cisco responded by denying access to router and switch operating system internals to the Spectrum developers at Cabletron, with the obvious result that customers using Spectrum to manage Cisco equipment were left out in the cold.
Vendor working relationships are the collective Achilles' heel of storage management software frameworks. That drawback would effectively be removed if the industry were to adopt a management standard such as SNMP or CIM. Instead, the trend is away from standards-based management and toward proprietary APIs, CLIs or Web page-based interfaces articulated by the equipment itself.
EMC and Tek-Tools
The latest entries in the storage management software business, EMC and Tek-Tools, approach the problems of storage management from distinctly different perspectives. Both are works in progress and may have changed substantially by the time this column goes to press, so be sure to check with the vendors about developments within their product groups before making any implementation or deployment decisions.
EMC, as stated last month, is endeavoring to externalize the management software used to control its own product families. The company is, in fact, offering its own robust API layer, called WideSky, as a de facto standard for integrating heterogeneous servers and storage platforms into a common management framework.
Despite what some critics say, EMC's approach is not just more annoying arrogance. By offering API access to storage vendors, the EMC approach reverses the pattern in framework software development in which the software guy had to write to all the hardware interfaces. In effect, EMC is offering immediate traction in EMC-dominated shops—of which there are many—to any vendor of point software or storage hardware who wants to sell into those shops. That's a powerful incentive that is sure to garner the interest of most storage players. In the process, EMC may grow its market share in the open storage management software market substantially.
Tek-Tools, on the other hand, is working toward a different goal. The company is currently following a somewhat conventional track in product development by developing management interfaces to storage platforms and point products as dictated by its growing customer base. However, the firm is also working with some interesting agent technology that will enable it to collect information from the self-articulated configuration and status Web pages articulated by storage devices—an increasingly popular approach among storage vendors—then normalize and display it in a uniform way on a central management console. What makes the approach interesting is that Tek-Tools doesn't require the cooperation of the equipment vendor to monitor or manage its hardware. The intelligent XML agent is potentially as much of a leveling force as any CIM or SNMP agent. Given the pedigree of Tek-Tools engineers in network management systems development (at Micromuse and elsewhere), they bring a fresh commonsensical "cool" to the endeavor.
Going forward, storage management software tools need to become more sophisticated to manage the diversity of platforms and topologies that are being deployed in modern enterprises. They need to provide automation of management tasks to reduce storage TCO and to realize return on investment objectives in storage acquisitions. Moreover, they need to add a new layer of reporting and integration capability to enable the delivery of storage as a service and to provide storage cost modeling information to business decision-makers. Seminal offerings in this space are coming to market from storage service providers-turned-management software vendors such as StorageNetworks and Storability.
Ultimately, the selection of storage management software is a personal one. Consumers would be well advised to consider the bottom line of improved storage management efficiency—and to test vendor claims—before they purchase. If Hannibal the Cannibal ever wanted a new profession, replete with dining opportunities, the data storage industry would be a good choice.
Jon William Toigo is an independent consultant, and a columnist and contributing editor for Enterprise Systems magazine. He's the author of 12 books, including Disaster Recovery Planning, 2nd Edition, and The Essential Guide to Application Service Providers, both from Prentice Hall. Contact him at email@example.com.