Friday, December 9, 2011

Part 13 - Insufficient data does not compute.

(This is part 13 in a series of 16 posts about IT leadership in higher education titled Everything I Need to Know about IT Management I Learned from Star Trek. See Part 0 - Introduction for the full list.)

Although insufficient data does not compute, a key management lesson from Star Trek is that often you must make decisions with insufficient data. You can’t move forward if you don’t make decisions. You will never have all the information, so you need to make decisions with enough information. Insufficient data may not compute, but as a leader, usually that is all you have to work with and you need to make the best of it.

The trick is to understand what is “enough”. As IT folks we are rational and logical so it is hard to make major decisions without all the data. For example, at my university we completed a brand new state-of-the-art data centre. Development of the business case was a challenge because it was a multi-million dollar decision. The existing data centre was out of space, power, and air conditioning because of explosive growth in research computing. Yet we wanted to develop a computing facility to attract new researchers to meet the university’s mission.

Before we could proceed, the board had two questions. First, why did we need a new data centre? Second, how big should it be? “Why” was easy. We had all the facts about why our existing data centre was too small. We drove the point home by giving the board a tour of the old data centre before we asked for any money. After the tour the board chair, who was non-technical, came out and was quite surprised. She said to me “I thought I just turned on my computer in the morning and it all just worked. I didn’t realize we needed all this. And by the way, it looks like you guys haven’t got anymore room!” That tour made the basic business case an easier sell.

“How big” was much tougher to answer because it was based on future predictions. Forecasting in a university environment has many problems. Research grants vary dramatically from year-to-year, so demand is unpredictable. Over long periods of time fundamental hardware technology requirements change such as mainframes to standalone servers to rack-based blades. Trends such as virtualization and cloud computing may dramatically alter what currently looks like a continuous growth curve.

With all that variability, our best guess said we needed about 5,000 sq. ft. over 5 years based on trends. After that, all bets were off. There is no way we had enough information to draw any conclusions past that time frame. But we didn’t want to build another data centre 5 years from now. So we did not have an air tight forecast. We decided to build a 10,000 sq. ft. building with one-half built as a data centre. The second half was setup as warehouse space that could be converted as needed to a data centre.

The result: in 9 months we filled 35% of the data centre. At that rate the built-out half of the data centre will be completely filled in 3 years, not 5 years! Because we included the additional warehouse space we have a contingency built in to handle what will likely be an overflow sooner rather than later.

We did not have all the information, it was a multi-million dollar decision, and we moved forward without being paralyzed by indecision. Based on the current utilization rate, we are glad we did. If we had not made a decision we would have lost significant amounts of research funding.

Insufficient data may not compute, but that does not mean you cannot make a good decision.


No comments:

Post a Comment