Deciphering TCO
Posted : Tuesday May 27th, 2008

TCO (Total Cost of Ownership) has been the in-vogue way for some time now to "objectively" compare the services and applications and the engines and/or operating platforms that host them. If you don't know about TCO then the basic idea is this. Server A costs $15,000 in up front hardware and software fees and then costs $250 a month in maintainence time. Server B costs $5000 in up-front hardware and software costs and then $500 a month in maintainence. So over a 4 year period the TCO for the two servers is as follows.

• Server A = $15,000 + (48 * $250) = $27,000
• Server B = $5,000 + (48 * $500 ) = $29,000

So what the TCO interestingly shows that while Server A has a higher upfront cost (3 times of server B) the total cost when all is said in done is actually lower for Server A then Server B over that four year span.

Seems interesting right? Problem is there are many flaws in TCO comparisons, some obvious and some not. Here are some problems I have with using TCO for comparison.

Difference in time spans

In the example above I arbitrarily chose to use a four year life span but the results are of course very different depending on the span of time one chooses. If you choose a 3 year cycle then Server A is actually the better investment and if you choose a 5 year or longer cycle then Server B increasingly becomes a better choice.

It would be nice of course to see TCO's standardized to all use either (or both) 5 and 10 year cycles because realistically in deployment terms most servers will see a usage somewhere in that range. Until then however as a discerning reader you have to make sure when comparing TCO's that you are comparing equal lengths of time and ideally lengths of time that are equal to the length of time you will be deploying the server or platform in general.

Unrealistic and/or unfair maintainence costs

Totalling up up-front and ongoing costs is a deceptively simple excercise because while up-front costs can be calculated to the penny, calculating ongoing costs relies on estimates, guesswork and assumptions and in the end just isn't as accurate.

There are three pitfalls in guesstimating ongoing maintainence costs that as purchaser you need to be aware of when comparing TCO.

1. Maintainence overhead - this one seems fairly obvious and most companies already take it into account in one form or another but it's still worth mentioning. Essentially ongoing maintainence for TCO when used for advertising purposes is based on a single installation but in reality you may have a number of existing servers of the same type that will be maintained by the same person or persons. There is a certain amount of overhead involved in maintaining a server that when spread out over more multiple servers becomes less costly.
2. Hidden maintainence costs - What costs are included in pricing ongoing maintainence and what are not? Maintainence costs are usually priced by the amount of time an administration person would have to perform work on a server in a given period of time. But are hardware costs included? What about systems that require regular upgrading through downloaded patches, is the bandwidth alottment for these downloads included in the cost.
3. Variance by service - Pricing ongoing maintainence costs based solely on a hardware and operating system platform basis is not wise because the amount of time and effort one can expect to take to maintain a server is going to depend heavily on the services it supports. For example a small database server sitting behind a corporate firewall is probably only going to require minimal maintainence with some allotment for emergency procedures. An email server that is on a public network on the other hand will require almost constant monitoring and tuning for both performance related issues and security concerns. Knowing exactly what uses a server is intended for gives you a much better picture of what the real ongoing maintainence and administration needs will be.

Marketing spin

The reality is that TCO numbers like other "hard" numbers come through the marketing spin cycle much softer than they started. While anecdotal evidence does not normally outweigh that of hard facts when the "facts" themselves are squishy to begin with anecdotal evidence should be added to the mix to try and get a better understanding of what the TCO for a particular product is.

Am I suggesting that people or companies are lying? No. What I am suggesting is that you have to recognize that what is being presented is a version of truth that best co-incides with the goals of the company or persons presenting it. As an example Microsoft often uses low TCO compared to Linux as a selling point for it's higher end enterprisey operating systems. In my own experience for example there is no question that while Microsoft OSes require far, far less training and experience for most administration requirements as opposed to equivalently functioning Linux based systems I know of several Linux based servers that have been rebooted two or three times over a span of 3 to 5 years whereas the Windows based servers that I am aware of usually require rebooting on at least a monthly basis. While my own personal experience is just that, a personal experience, it is interesting to me to note that depending on which version of truth you want to use it either supports or contradicts the idea of a lower TCO for Windows.

Tags

measuring  objectives  TCO  uptime 

Categories

Technical  Business 

Comments