Download the authoritative guide: Enterprise Data Storage 2018: Optimizing Your Storage Infrastructure
By MacFadden's calculations, an SSD array capable of delivering at least 250,000 IOPS would cost around twice as much as an HDD array offering the same performance. That's even allowing for the fact that to achieve similar performance you would have to buy some five times more HDD than SDD capacity, because short stroking would restrict utilization to about 30 percent on the HDD. (The SDDs could probably achieve 70 percent utilization, even in a RAID 6 configuration.)
He then estimates that some additional costs would be about the same—provisioning, administrator training and even warranty and maintenance costs.
But due the difference in the number of disks, some of the costs will be very different indeed. Storage admin costs for the SSD storage array would be 20 percent that of an HDD array. Power and cooling would be about 10 percent. Facilities (such as rack space), just a fraction over 10 percent.
That means that over three years the total cost of ownership of a high performance SSD array would be about half as much as an HDD array, even though the initial purchase price of the SSD array would be twice that of the HDD array, according to MacFadden's model.
Structurally High TCO
Some organizations have a structural propensity to purchase systems with a high TCO when lower TCO alternatives exist, MacFadden warns. The reason comes down budgets, and the way they are allocated, he explains. The buying decision for a storage array is often made by the IT team, and the purchase cost comes out of the IT budget. But many of the costs of owning it—power, cooling and floorspace, for example, are borne by the facilities team and come out of another budget.
As a result, there is no reason for the IT team to buy an SSD-based array if an HDD-based one that offers similar performance can be purchased for half the initial outlay. But for the facilities team, this could mean using up five times as many racks and consuming five times more power and cooling resources. For the organization as a whol,e that means a TCO that's twice as high as need be.
It's only when the buying decision takes the TCO in to account—not just the hit to one team's budget—that the best purchase for the organization will be made.
Low TCO Features
When considering TCO, one obvious question is bound to come up: are there any general storage system characteristics or features that promote lower TCO?
According to a Storage Economics study carried out by Hitachi Data Systems in 2009, certain storage architectures do have a built-in TCO advantage. The study concludes that key ingredients for lower TCO include
- Virtualization of volumes, file systems, storage systems
- Dynamically tiered storage
- Intermix storage (disk types) within the same storage system (solid state disk or SSD, SAS, SATA)
- Thin provisioning
- Power down disk, Massive Array of Idle Disks (MAID)
- De-duplication, data compression
- Integrated archive
- Capacity on demand
- Policy-based storage provisioning
Storage management software that provides higher utilization rates or boosts performance—or both—can also play a role in reducing TCO.
Examples of this include VMware's Virsto storage hypervisor, which handles I/O more efficiently than VMware's server virtualization hypervisor, and resource allocation software such as VMTurbo's Operations Manager, which uses a market-based system to allocate storage resources efficiently by giving applications "budgets" to use to purchase storage.