By Jon William Toigo
In April 2005, the Storage Networking Industry Association (SNIA) held a press conference to announce the results of a conformance testing program focused on its Storage Management Initiative–Specification (SMI-S). SMI-S is based on two existing standards from the Desktop Management Task Force–Common Information Model (CIM) and Web Based Enterprise Management (WBEM)–that SNIA intended for use in managing heterogeneous storage infrastructure.
Last year, Vincent Franceschini, of Hitachi Data Systems, and the new SNIA chairman, declared the SMI-S development effort a “mission accomplished.” SNIA’s work had been approved by the International Organization for Standardization (ISO) and the International Electrotechnical Commission (IEC) as a standard for interoperable storage management technologies. Moreover, the chairman said, SMI-S was now implemented on approximately 450 storage products.
For perspective, in 2006 alone, more than 11,000 new storage products were introduced. Simple math suggests that SMI-S, hailed as the re-invention of storage management technology in a unified and standards-based way, has not caught on in the storage mainstream.
There have been many theories to explain the outcome–from the daunting costs associated with designing a product for SMI-S based management, to the politicking behind the development of the specification (which some claim froze out many smaller vendors), to the failure of SNIA to promote SMI-S effectively outside of its own members. In any case, SMI-S hasn’t delivered the fix for unified storage management, and this is unlikely to change in 2008.
SMI-S aside, the key impediment to the development of any universal storage management solution is the reluctance of storage vendors to commoditize their wares. “Value-add” software, joined at the hip to a proprietary array controller, is the most common path to product differentiation, as well as a means for vendors to lock in the consumer and to lock out competitors.
One problem with these value-add components is that they too often make getting an accurate read of available capacity, data pathing, and other information essential to proper storage resource management very difficult. Some vendors reserve capacity on their arrays for software they want to sell the consumer now or in the future, but they do not provide tools that the customer can use to see how much of their capacity is reserved but not allocated. Result: Determining how much of the available capacity is actually usable for the business is an extraordinarily taxing effort.
Conversely, many vendors have applied “thin-provisioning” techniques to their arrays, which makes space that is “reserved but not yet used” by applications available for other apps and end users. Thin provisioning is a high-tech shell game that is only as effective as its underlying demand forecasting algorithms. Should the application that “thinks” it owns disk space that has been thinly provisioned for other purposes ever call for that space, chances are good that the application–and the server–will fail. As with vendor capacity reservation, thin provisioning obviates a clear view of exactly how much capacity is available at any time.
Also limiting the ready assessment of storage capacity are proprietary data-pathing schemes that appear in the arrays from many vendors. Proprietary data pathing is often used to increase the throughput of an array by establishing multiple connections to the same volume, LUN, or disk drive. Unfortunately, this strategy can also result in the erroneous re-counting of the same capacity as seen through each path, again obfuscating a clear view of how much real capacity is actually available.
What mainly passes for storage management these days is specialized on-device management software. Device-level management is not regarded as a problem in shops that have an IT staff of only one or two people. In fact, these device-level managed products, from companies such as Compellent, 3PAR, Dell/EqualLogic, and Xiotech, have demonstrated significant appeal in the small to medium-sized business (SMB) space.
But even smaller businesses grow over time, and their storage platforms become more diverse and heterogeneous. As such, managing each storage platform using its own command-line interface or Web page can become more time-consuming and labor-intensive.
Hence, the first evolution of distributed storage management has been the development of “horizontal capacity managers” that aggregate, and in some cases augment, device-level management approaches. These storage managers, called storage resource management (SRM) software suites, access gear using, among other things, vendor-exposed application programming interfaces (APIs).
CA, IBM, Symantec/Veritas, and Tek-Tools are among the lengthy list of providers of SRM products. SRM product managers say that creating and maintaining SRM tools requires an ongoing investment by the software vendor to license access to proprietary APIs from each storage platform vendor, and then undergo the additional hair-pulling exercise of validating and certifying their wares with each vendor. This approach is expensive to SRM vendors, which must support the range of storage devices most commonly used by customers, and to the consumer, to whom the costs are ultimately transferred.
In the next part of this two-part column, we will look at what SRM vendors think about unified storage management and examine the evolution of SRM software.
Jon Toigo is CEO and managing principal of Toigo Partners International LLC (www.toigopartners.com).
The evolution of storage management
- Device-centric management (where most companies are today)
- Horizontal/infrastructure-centric management (what most SRM vendors are striving for: heterogeneous equipment support via a combination of API interfaces, SNMP MIB data, software agents and, where available, SMI-S)
- Application-centric management (what we need to get to in order to support troubleshooting/tuning of application-to-storage connection–an SRM product that provides an application- or business process-centric view of infrastructure). Sets the stage for the final evolution of storage management: data-centric management or information lifecycle management.