A recent article in Forbes extensively quotes a study by Forester Research to identify four widely used techniques -- tiering storage, data retirement, compression and storage virtualization -- for "storage capacity growth containment."
Technically, tiered storage has nothing whatsoever to do with capacity growth containment. It's a strategy for data movement intended to primarily free space on more expensive and higher-performing storage by moving less frequently accessed data to less expensive and lower-performing storage over time.
Storage tiering requires some infrastructure support: You must have both high-performance and lower-performance storage, the latter usually providing high capacity. As readers of Forbes would probably agree, it's best not to have your entire investment portfolio in a single economic sector. So too with storage tiers: Tape has always played the role of tier-three storage because of its extraordinarily high capacity and low per-gigabyte cost, while we've recently seen flash SSD suggested as a tier-zero layer because of its super-fast performance, very high cost and comparatively low capacity. In between these layers, we see tier-one storage comprising 15,000 rpm Fibre Channel (FC) or SAS drives and tier two representing lumbering, high-capacity, low-cost SATA arrays.
Tiering does nothing to bend the capacity-demand curve, though it may slow the rate at which tier-one storage is purchased. According to Fred Moore, president of Boulder, Colo.-based Horison Information Strategies and whose estimates on data storage performance are widely quoted, a balanced tiered-storage infrastructure will see between 1% and 3% of data stored to tier zero at $50 to $100 per gigabyte. This data will migrate over time to tier one, where it will join roughly 12% to 20% of all data on high-performance disk arrays costing between $7 and $20 per gigabyte. When access frequency to this data declines, it will migrate to tier-two storage, which hosts between 20% and 25% of business bits for $1 to $8 per gigabyte, by Moore's calculations. Then the data will move into tier three, where between 40% and 60% of corporate data should reside, renting space at 20 cents to $2 per gigabyte.
Unfortunately, as Moore points out, many companies content themselves with only two tiers, dramatically reducing the cost savings of tiering. For a 100 TB storage complex, Moore argues, the cost for hosting data using only tier-one and tier-two devices is approximately $765,000. Adding a tape tier drops the price to about $360,000 and, according to Moore, adding a tier-zero layer takes the price back up to $482,000.
Those numbers don't suggest that tiering storage in your environment will reduce capacity demand. Well-planned tiering, however, can forestall the need to grow the top tiers of the storage pyramid and save money ... in theory. The truth is that a lot of vendors have been blurring the lines between tier-specific products in their storage-product portfolios. Tier zero and tier one have been merged in some cases to provide hybrid flash arrays, and tier one and tier two have been mixed in with other products in an effort to push tier three out of the pyramid. The resulting products are advertised as making tiering storage simple by automating it within the firmware of the array controller, reducing the labor cost of arrays. This may or may not be true; the real question is whether the labor-cost savings offset the huge markups on commodity flash chips and/or FC, SAS and SATA hard disks that usually accompany such offerings.
Whether you realize the benefits of tiering as touted by Moore and others also depends on the degree to which storage purchasing is a rational act. In a recent IBM storage webinar, a company spokesperson noted that too many customers buy too much tier-one storage: "With every new application they roll out, they buy the latest tier-one storage array to host its data." He wasn't complaining, just observing.
There is a glut of tier-one storage deployed today, and given all the excitement around this new phenomenon called flash, most analysts expect more of the same. Meanwhile, the appeal of tape, even with significant and compelling advances in capacity and format, seems to be diminishing. But if tier three disappears, the cost-benefit appeal of tiering your storage will hardly be worth the effort.
This was first published in May 2013