* Secrets of running a data center well
We have spoken at length in this forum about the challenges of managing data centers, both as technological ecosystems and as physical entities, in the face of explosive growth in storage and computational demands, heat generation, and power consumption. But does it have to be like that? Does the data center have to experience change of this sort? If not, what are the factors you can change to help avoid the changes, rather than simply coping with them better?
The most successful data center I ever saw was anomalous in several ways that I think are worth considering. First, and most shockingly, it was much too large for the equipment it needed to contain. It was not brand-new and yet to be filled, either. Instead, it had been carefully managed so that as servers became smaller and more powerful, the need for space shrank. Nothing in it was new for the sake of newness, nothing in it was upgraded, expanded, or replaced without a solid business case, and when things were replaced, they were replaced – that is, the old system left after the new one came on line. Some commodity activity, like Web serving, was outsourced. The option to scale up was explored before scaling out – one beefier box instead of 10 cheaper ones. Consequently, it had no space, heat, or power problems.
Secondly, it was not heavily trafficked. Admins did not – because they did not need to – spend much time in the room. They had a small but versatile set of management tools that let them do most of what they needed to from their desks. They also had a strong change management process that minimized the number of surprises they gave each other.
Thirdly, few of the people working in that data center were new to it. The average tenure of the small crew of systems administrators was about 10 years, and much of the efficiency of management derived from that fact. These admins knew their systems, their tools, and their vendors and consultants. They also knew each other, how to work with each other, and had enough cross-training and learning-by-osmosis to be able to back each other up. The measured and rational approach to change meant that these admins generally had time to learn the new skills they needed as they needed them, and that new staff brought in had time to get oriented to the way things worked there and to share their own new skill sets and perspectives.
This intertwined set of characteristics, it seems to me, centers on human factors. The people leading IT at that company and running that data center understood and resisted the very human love of change for change’s sake. They understood and capitalized on the human capacity to build deep expertise as well as broad but shallower competencies, and the fact that the synergies in a tightly knit group of managers can make them much more efficient than a bunch of folks who don’t know each other. They also understood that retaining people made this kind of environment easier to create and maintain, and at the same time that having that kind of environment made retaining people easier.
I might sum up an underlying theme of many of these newsletters as, “Manage your environment rather than letting it manage you.” It would be wise to look at how that extends to the human environment as well.
Editor’s Note: We regret to let you know that we will no longer be publishing the New Data Center Strategies newsletter. The last newsletter will be mailed on Aug. 7. To keep you abreast of data center news, starting the week of Aug. 13, you will begin receiving the Data Center News Alert, which mails every Tuesday. We thank you for supporting Network World newsletters.
