The cost of storage media is plummeting: both hard and solid state drive capacity has never been so inexpensive.
That would be great news for companies that have significant storage requirements, except for two inconvenient facts. Firstly, the purchase cost of a storage system is made up of much more than the cost of the storage media itself. And, more importantly, the purchase cost of a typical hard disk drive (HDD) based system accounts for less than 20 percent of the total cost of ownership (TCO), according to Parity Research.
So while the purchase price of a storage system is evident at the time that you hand over money to a vendor, it is not the whole story. What's important is the TCO over the lifetime of the system—typically about five years.
Some of the post-purchase costs are "hard" costs like power and cooling, which are easy to calculate up front. But there are also "soft" costs, which are more difficult to pin down and which often get overlooked. These soft costs include the cost of taking on risk—such as the risk of downtime or, less obviously, the risk of non-compliance with regulatory requirements.
"I think that a lot of folks that buy storage dramatically underestimate these soft costs, particularly the risk of downtime," says Gary MacFadden, the founder of Parity Research.
Some of the most significant hard costs over the lifetime of the system include the following:
- Procurement costs
- Hardware and software maintenance costs including extended warranty costs
- Management staff costs during the life of the system
- Monitoring system costs
- Power, cooling and backup power supply costs
- Data center space costs
- Backup infrastructure costs
- Disaster recovery costs (including disaster recovery service subscriptions or maintaining compatible storage at secondary sites)
- Storage area networking costs
- Security costs
- End of life costs including the cost of migrating data to a new storage resource
In addition to the risk of unscheduled downtime, mentioned earlier, other soft costs that should be taken in to account include the following:
- Cost of waiting for storage to be provisioned
- Unused storage capacity cost
- Cost of storing duplicated data
- Scheduled outage costs
- Risk of lost data
- Poor storage performance costs
High Performance TCO
This last cost—the cost of poor storage performance—can be particularly significant in large enterprises with customer-facing applications, says MacFadden. "Some applications are all about serving customers, and if the screen doesn't refresh in 100ms the company could lose business worth $1m a day," he says.
Solid state drives (SSDs) offer far higher performance than HDDs, but they are also more expensive per gigabyte than HDDs. So if you are purchasing a storage array solely to provide extra storage capacity, the TCO of an SSD based array is bound to be higher if you assume that all the additional costs are about the same. (In actuality, you could argue that SSDs are more reliable and consume less energy, but let's ignore that for the moment.)
But if high performance storage is the main objective, then the economics, and the relative TCO, change dramatically. "If you are after high performance for a SQL app or an Exchange server, then you can use hard disk drives, but you need to buy many times more HDDs than SSDs to beef up the IOPS," says MacFadden. That's because enterprise grade SSDs may offer 4000 IOPS, while even the highest performance 15k rpm HDDs only offer around 400 IOPS. To get equivalent performance from an HDD-based array requires short stroking—a practice which only utilizes a small proportion of the capacity of a large number of drives.
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.