Published: 10 Aug 2006
How to count the cost of storage
If you want to take some serious steps toward saving money on storage, you need to figure out how much things truly cost.
A KEY FOCUS for storage professionals (and information technology in general) is saving money. Although we work in a field with rapidly declining unit prices, real cost savings have been difficult to prove. Even as vendors tout the cost effectiveness of new storage systems, and high utilization and tiered storage become a reality, storage-related costs seem to keep climbing. Perhaps the problem is that we're focusing on the wrong areas and measuring the wrong things.
Supply and demand
Let's get this out of the way first: The cost of each gigabyte of storage is declining rapidly in every segment of the market. Enterprise storage today costs what desktop storage did less than a decade ago. So why are overall costs increasing?
Part of the blame lies in our voracious appetite for storage. We may cite rich media for creating ever-larger files, but that's not the culprit. Even the constant flood of e-mail isn't the prime suspect. In fact, most storage growth is the result of low utilization and multiple copies of data--our inability to make good use of the disk already deployed. In effect, we're drowning in empty disks!
The average enterprise today uses just 10% to 15% of its raw data center storage. This equates to 20% to 30% of usable disk space once necessary elements like RAID, replication and sparing are taken into account. This number has actually fallen somewhat since I wrote about storage utilization back in 2003 (see "Real-world storage utilization," Storage, April 2003).
The low-utilization trend is hard to stop. Migrating servers and data to a new model is incredibly complex, risky and time consuming. For this reason, many businesses have decided, after a utilization review, to ignore the problem in existing systems and focus instead on new allocation. But even if a company institutes radically more efficient processes for allocating storage, it would take years to refresh enough of the estate to make much of a dent in the overall utilization rate.
Of course, there's justifiable growth in storage demand. But duplication is another factor. Mirrors and replicas were once exotic techniques used only for the most critical data, but they're common today. It's not unusual for a file to be mirrored for redundancy, again for backup, a third time for testing or decision support, and to then be sent offsite for disaster recovery. One company I know routinely keeps six copies of most production data online and dozens more on tape.
Combine low utilization with natural data growth, multiply it by duplication and it's an equation for disaster.
Total cost of service
Anyone who has worked in financial accounting can tell you that unit cost is just one small metric, and the total cost can often be much higher. In storage, it's critical to take a broad view of the actual cost of delivering a service rather than the unit cost for a gigabyte of storage. In other words, what else goes into the system apart from disks?
The most obvious cost elements are hardware and software. Then add in data protection in the form of backup software and tape hardware, and disaster protection with replication software and hardware, extra storage arrays and so forth. And few businesses would dare to run their systems without maintenance, which means support contracts that can easily add an extra 30% to the overall cost of storage.
Stepping back further, there's the infrastructure that allows data storage to exist. Data center costs can add up quickly, especially with rising energy costs. The cost of consumables, such as tape media and offsite storage, is also often overlooked. Some organizations even have "keep forever" policies for tape media, causing these costs to rise significantly.
Of course, personnel-related costs are the elephant in the room. It can be extremely difficult to accurately allocate management costs into a cost calculation, especially for employees who aren't 100% dedicated to storage. But one of the biggest issues with personnel costs is the conundrum of value: It takes little effort, and thus little cost, to do a poor job of managing storage; it takes much more money and time to do a good job. In fact, whenever a real assessment of storage management effectiveness is performed, it's unusual to find a company that's overstaffed.
Note, too, that the ratio of these cost areas isn't consistent from one company to the next. Small businesses tend to spend much more money on capital hardware and software, while larger businesses are more personnel heavy. In fact, a TCO analysis at a large enterprise may reveal that hardware and software account for a very small percentage of the overall cost of storage--sometimes even less than 10%.
More than one way to count the costs
Just as there are many metrics that go into cost analysis, there are many different (and equally valid) ways to add them up. Businesses are typically concerned with two types of cost analysis: cash outlay and balance sheet impact.
Let's look at these in terms of a car purchase. A cash analysis would begin with the cost of the vehicle and the sum of the finance charges and fees. That's how much money is required to take ownership of the asset. Then you might determine your operating costs by calculating the cost of fuel, insurance and taxes, and maybe some factor for wear and tear. A cash analysis for storage is similar. Calculate the total hardware, software and installation cost, and add in finance charges to determine the cost of acquisition. Then calculate the cost of ownership by adding in items like data center, personnel and consumables for your estimated ownership period.
A cash analysis is useful to compare competing systems, but it has little to do with financial realities. Using the car purchase analogy again, think of a balance sheet (or P&L) impact assessment as a comparison of leases and monthly payments. How much will this car (or storage system) cost next month? Most durable equipment is either leased, financed or depreciated. The end result, and method of calculation, is similar: Add up the acquisition cost and fees for a given period and divide it evenly. Depreciation of a purchased asset is generally straightforward, while leases can be maddeningly complex. But again, there's more than the acquisition cost to consider. The periodic cost of ownership must also be taken into account for an accurate cost assessment.
One surprising aspect of cost analysis is that the cash and P&L analyses don't necessarily equate. It's likely that some items will be used beyond their depreciation threshold, effectively creating "free" equipment and causing the P&L to look awfully attractive. On the other hand, vendors can often sweeten an acquisition deal to swing the cash analysis in their favor.
Cost avoidance vs. savings
I began this column by talking about the great pressure to save money. But how can we save money when the total cost is rising? There are two ways to save money: cost avoidance and cost savings. The former means that you avoided an expenditure that seemed inevitable. A true cost savings would require overall costs to drop, a much more difficult proposition as data demands expand. Luckily, even a documented cost avoidance will be sufficient to show that we in the storage realm are doing our jobs.
How will you count the costs? If you're trying to compare devices to each other, I suggest performing a cash analysis over your expected period of ownership. The result will be a statement like "It will cost X to buy this device and run it for five years, while that one will cost Y." Choosing the less-expensive option will count as cost avoidance because you had to buy one of them.
On the other hand, if you've been asked to come up with a chargeback metric for storage, it's preferable to use the P&L impact method. Consider only the actual costs of delivering storage, including the depreciation or lease cost, allocated-in personnel and data center costs, and other bills that are assigned to the storage part of IT. Add up the actual cost of delivering storage in general and divide it by the total amount of storage provided to come up with a fair allocation of cost.
No matter where your cost analysis leads you, don't rely on "industry-standard numbers." In my years of performing cost analyses for a broad array of companies, I've found that there's no standard for the cost of running storage. Stick with your actual costs. Get out your hardware, software and maintenance invoices, chargeback reports and time sheets and add everything together. Nothing will impress the bean counters like a real accounting of actual company data.