Does ‘a little extra’ really go a long way?

Maybe there really is a reason you need to drive a Hummer or some other mobile behemoth for the daily commute to work, and perhaps you were right to believe the salesman when he said a $1,000 fountain pen really does spell better.  Maybe. Perhaps you can get away with that at home, but these days there is just no way to justify frivolous expenditures for storage.  Hence this week’s rant.

When it comes to buying extra disk drives and the cabinetry to go with them, we have a bad habit in IT of buying what we think we need … and then buying a little bit extra, just in case. 

The “just in case factor” is to cover our nether regions just in case demand for primary storage rises faster than our ability to bring more on line.  We usually refer to this as “headroom,” although our accounting departments probably have another term for it.

As you are probably aware, Anne and I have a strong prejudice in favor of storage resource management (SRM) software as the way to get your storage assets under control.  We point out whenever we can that an investment in the right SRM solution is the most economical way to squeeze all the performance and space utilization as possible from your site.  It is also a way to make your support team more productive as the more sophisticated offerings provide automated policy-driven responses to a wide variety of situations.

Eventually, of course, we all have to buy more hardware.  At such times, it is useful to keep in mind some basic rules of economics. 

By way of example, last week I had an interesting discussion with an IT director.  He runs a good-sized shop, numbering among his resources a mixed bag of storage devices that includes some big iron from one of the major vendors.  He is reasonably happy with what he has bought in the past, the field service is good, and he is in the process of buying more arrays.  But how much should he get?

The answer probably is to buy just what he can use today and no more, with no attempt to plan for headroom.  A look at the pricing history of the array market justifies this.

In the old days – the mid-1990s – disk drives often accounted for more than 50% of the price of a fully populated RAID device.  That is all different now, of course.  We all know that the price of a gigabyte of storage has dropped precipitously since then, and that it has done so at a more-or-less predictable rate that seems to result from yet another application of Moore’s Law.

Moore was of course talking about the annual doubling of transistors per square inch in an integrated circuit, but the law seems to apply (albeit at a slightly lower rate) to increased areal density on an HDD as well.  That, plus cutthroat competition, has meant that only those disk drive producers that could live with very short product life cycles and razor-thin margins have survived.

The result of this is the price of a $2,000 disk drive today is likely to be cut by 40% next year.  

It is important to point out here that the annually published costs of a gigabyte of storage are of limited value in this context.  High-end arrays typically only use disks with the highest spindle speeds available (10K or 15K rpm), and the array vendors charge a steep premium for pretesting them and doing a burn-in on the devices they ship you. 

I suggest you ignore the specific cost-per-gigabyte numbers (which in most cases only reflect the mass market drives) but do make sure you understand the trend they indicate.  High-end drive prices follow this trend, although at a somewhat slower rate.

A simple return on investment analysis will show that it makes little sense to invest in drives before you need them, and almost no sense to buy more cabinetry than you currently need unless you get one heck of a good deal and the maintenance charge is at the noise level.  For example, a $100,000 cabinet populated with 50 drives at $2,000 each will cost $200,000 today.  What would that cost next year, when you will be paying with cheaper dollars?

If you don’t believe me, ask your friends in the accounting department about this, and don’t forget to ask them about the concept of net present value as well.  You will be pleasantly surprised.

Maybe a new rule will be “never buy today what you can put off purchasing tomorrow.”  After all, the dollars will be cheaper, and the devices are likely to be cheaper as well.