When taking the long view of storage evolution and surveying the types of storage products vendors have delivered to market, you might get a sense that consumers are a bit, well, schizophrenic. On the one hand, we crave more storage capacity, which is achieved these days by adding more and more boxes of disk drives to shared switched fabrics -- what we call SANs. The chief goal is to minimize the possibility of a catastrophic "disk full" error message that takes down a mission-critical IT application.
On the other hand, data storage cost is still a key consideration. The "do-more-with-less" mandate tends to seep into the IT purchasing process of even well-heeled firms. In response, vendors are quick to introduce compression, deduplication, thin provisioning and other software-based technologies to enable consumers to store more bits on existing spindles (or, in the case of thin provisioning, to defer spindle purchases until they're absolutely necessary).
This rationale for storage purchasing -- to keep up with the capacity demands of growing data stores and to restrict expenditures where possible -- leads to an equally schizophrenic definition of data storage efficiency: namely, that storage efficiency is a measure of how well you meet capacity demands with the lowest possible budgetary expenditure. In turn, we're encouraged by vendors to believe that true storage efficiency is derived from value-added software sitting on top of commodity hardware.
But why aren't we paying less to acquire storage if the cost of commodity hardware is falling? The answer is that the cost to license the software is dwarfed by the cost of the hardware. For example, 300 TB of SATA hard disks, with an estimated street price of $3,000, has an MSRP of $400,000 because the vendor adds some deduplication software joined at the hip to the array controller.
Consider data storage cost and longevity
So, how do we strive for smart storage investments that result in greater storage efficiency? Here are two basic things to consider:
Your infrastructure components will always be a generation, or two or four, behind the latest technology.
First, the technology should survive without becoming obsolete for five to seven years. That said, a buyer must assume the product being acquired will be declared end-of-life by its vendor within 17 months of its introduction to the market. This means your infrastructure components will always be a generation, or two or four, behind the latest technology. This doesn't mean your storage is obsolete, only that buying the latest technology today doesn't deliver any sort of continuing advantage over a competitor. The best you can hope for is that your storage purchase will continue to find productive use for the number of years that you wish to use it.
Second, the data storage investment should include a known set of recurring and operating costs for each year of its useful life. In addition to leasing/financing costs and energy costs to power and cool equipment, you need to anticipate that at least one in every 10 disk drives in the array will likely need to be replaced annually and that, most likely, you'll need to purchase replacement disks from the original array vendor, who charges a very high mark-up for "signing" the disk drive. You also need to anticipate that the rig comes with a warranty agreement that's good for three years, which is a few years shy of the useful life that you require of the hardware. Upping the warranty to cover an additional two to three years of maintenance for hardware and software components will likely cost an amount equal to or greater than the purchase price of an entirely new rig. All these factors need to go into your calculation of the annualized cost of ownership.
But it's a challenge to calculate and quantify what value the storage gear delivers so it can be compared to operating costs. In addition to quantifiable value, there may be non-quantifiable benefits that accrue to the product acquisition, such as the following:
- Does having a well-known vendor's brand name on the bezel plate of the storage array deliver any value? Some cloud storage vendors would say it does, claiming customers prefer to use their service because their data is stored on "enterprise-class" storage arrays as opposed to "white-box" arrays (despite everyone selling the same box of component parts).
- Similarly, the purported compliance of a storage product with legal and regulatory mandates that govern data protection, retention or privacy may provide an intangible incentive for some consumers. (Note the word purported. Many vendors offering "compliant" products neglect to tell the consumer that there's no governmental entity certifying the compliance of any hardware. Such claims should be a red flag.)
- The vendor's stock price or longevity in the market may encourage loyalty or influence purchasing preferences.
Collectively, these intangibles may add to the quantifiable value received from a storage investment, offsetting its operating costs and providing an annualized estimate of ROI that helps define storage investment efficiency.
Investment efficiency may ultimately come to be the most important definition of data storage efficiency, since it's the lens through which corporate often views storage. Indeed, in more and more of the companies I visit, the rationale for storage acquisitions is couched in the terminology of investments and ROI rather than scientific measures preferred by engineers or process-oriented metrics preferred by operators. In a few cases, storage is still purchased simply on the basis of cost per terabyte.
What's most likely to be missing in all methods for measuring storage efficiency is any sort of attention to storage management -- be it managing the data placed on the storage infrastructure or coherent common management of the heterogeneous arrays deployed by a company. Without effective storage management, you need to add more humans to administer infrastructure, thereby increasing labor costs for storage. Without effective management, you're less likely to be proactive in your responses to burgeoning equipment failures (assuming you even have a clue they exist). So, uptime is jeopardized by a lack of management, as is ROI predictability.
In the final analysis, storage efficiency remains a poorly defined standard. But, given the increasing slice of the IT hardware budget pie being allocated to storage products every year, storage efficiency remains a standard toward which most IT departments claim to strive.
Realizing an efficient storage operation requires consideration of not only the functional attributes of storage technologies and their acquisition and operational costs, but also their interoperability with differing workloads and other heterogeneous infrastructure components. Most important, however, is to manage storage at the hardware and plumbing layer (via storage resource management) as well as at the services layer (via storage services management).
Organizing a storage practice without a clear definition of what constitutes storage efficiency, and without a clear reference for how storage gear and services will be managed, would be better termed a storage efficiency avoidance strategy.