After many years in this industry it came as a surprise to me that more than 90 percent of the total cost of ownership of a data center is locked by the time you issue the construction contracts.

As an engineer by education I found it hard to believe that I could ‘optimize’ during the building’s operating life to my heart’s content, and still only impact the TCO by less than 10 percent!

Given then how little you can tinker with a data center during its operating life it amazes me how little attention is paid to TCO during the design phase of a new site. In my line of work I get to see hundreds of new data center designs at all stages, from early concepts, through to final designs that have been compromised after the ‘value engineering’ phase.

Think about it early
Every time I read the news about the next 'mega data center' build being announced, especially those built in the public sector, my heart sinks a little, as I know very little real financial thought beyond the first capital spend will go into the design process.

Of course the folks involved will say otherwise, that they do consider the in-operation efficiency and cost, etc but realistically it’s a token gesture and not an investment grade financial analysis. And guess what folks, these data centers are massive investments, so the financial due diligence that should be happening on them just isn’t in over 90 percent of today’s market.

All data centers are financial investments in their own right and thus need to perform as such. Typically when the budget is being set, everyone is focused on the initial capital budget required and unless you have unlimited capital, minimizing the initial capital of course should be a very real goal of the project.

However, trying to minimize the initial capital without also calculating and understanding the total cost of ownership, will inevitably mean you will make decisions that compromise this total cost of ownership. If you do that, you’re effectively compromizing the return on investment you ‘could have had’ on the overall investment.

Unless you fund and then build a data center that you immediately flip (sell), your investment term will go well beyond the first capital and over time be more and more influenced by operating cost and future inevitable reinvestments. Remember, data centers have a lot of capitally intensive equipment whose viable operating lifetime is considerably less than the building’s shell itself!

Designers don't look at operations
Part of the problem here is that often the ‘designer’ isn’t motivated or incentivized to look too closely at in-operation costs or total cost of ownership, and the person who ultimately has to live with the facility for the majority of its operating life has no input.

Worse still, trying to manually assess all the possible choices that result in a Capex-vs-Opex tradeoff during design time is both complex and time consuming and thus costly. Of course the relative cost of doing this type of assessment at design time is minuscule compared with the resultant cost savings achievable from the results, yet so many people either don’t do it, just treating it as ‘another’ cost that can be cut, or think they can somehow divine using their years of experience and ‘rules of thumb’ that have always supposedly served them well.

Seven or eight years ago I would have had no issue with using experience and 'rules of thumb', but after the intervening period designing and using software to undertake that type of analysis; when we used that software it produced the sort of counter intuitive results that proved our preconceptions and experience to be incorrect.

As the years went by, time and time again, real analysis proved our own and others expectations to be wrong, I began to realise that I really shouldn’t be so surprised. Data centers may look simple - a building with an electricity supply, some UPSs and generators and air conditioning to cool the space - but in reality they are increasingly complex. Over the last ten years they have moved from designs and technology that were very static and stable in operation, to now highly variable and much more 'unstable' in operation.

Today’s data center has become a mass of competing and dependent variable sub-systems, dealing with a much more dynamic load and often much more sensitive to the climate they are built in and thus need more complex controls.

This is both good and bad news from a financial perspective. The good news is they have the opportunity to be much better at their job - delivering uninterrupted power and cooling at minimum cost - the bad news is that designing one that actually works is much harder. Doing the financial analysis on how it will or should perform is harder still, and consequently the result can be a disappointment both to the technical staff that need to operate it as well as the folks that spend the money to build and operate it.

So, next time you or someone you know even contemplates spending millions on a new data center (or even a sizeable reinvestment into an existing one), ask if the words ‘Total Cost of Ownership’ are part of the discussion. Ask if the people designing it can guarantee the performance of what it is that they are designing? If the answer to those two questions is ‘no’, then my friend, may the Force be with you, because you may as well be playing a game of ‘luck’ than a game of ‘skill’.