Strategy should precede tactics, however lessons from tactics often spark new insights and refinements to existing strategy. This is the case with data center power consumption, where our work over the last few years now points out a need for us to widen our focus if we’re to move computing to a truly sustainable model.

At Dell, our research has convinced us that the industry needs to move in a different direction – one in which power and cooling architectures are still important, but where IT organisations take responsibility for power consumption.

Our modelling work shows that suboptimal IT deployment and operations decisions directly result in constrained data center capacity and data center run-life. Before raising the alarm with facilities for more power, more cooling or more space, IT has to look at itself as the first responder. Energy smart policies with respect to utilisation of IT equipment and appropriate refresh of hardware equipment can have profound implications on data center energy consumption.

Reflecting on this lesson, I was struck by how well the concept of refresh applies to all aspects of our current situation. Not only do we need to refresh our servers, but we also need to refresh our tactics, our strategy and our understanding of the issues involved.

The EPA’s final report to Congress on server and data center power consumption was released in August of 2007, recently celebrating a year and a half in the public eye. Much of the data supporting the document was gathered and analysed in 2006. I use, and see, the EPA’s chart on historical and future energy use almost daily.

Now, with the supporting data two years old, I’m starting to get asked: ‘Well, where are we? Have we followed the path defined by the Current Efficiency Trends scenario? Are we above? Below? Do we see a path that leads us to the state of the art curve? How have we done on overall server power consumption? Do our original methods for calculating server and data center power still hold? Do we have better data today than we had in 2006? Better methods?’

The EPA’s report to Congress is full of information – descriptions of barriers to energy efficiency, solid data and lists of policy recommendations. Some of these are still relevant, while others may be obsolete. It’s time for a complete review, not only of the assumptions and data used as input to the report, but also of the insights, conclusions and recommendations embodied in the document.

The EPA’s report to Congress was also used to create provisions within the Energy Independence and Security Act of 2007. Some of the initiatives directly or indirectly driven by this act were a subsequent EPA request to participate in an industry data center study, the National Data Center Energy Efficiency Information Program, development of the DoE’s DC Pro tool for data center assessment and the DoE’s data center pilot programme.

The list of programmes and initiatives is impressive, but did we make the right decisions, did we focus on the most effective activities? Close to two years later, do we need to amend, add or adjust these provisions?

My favourite piece of the EPA’s report is Section 7.1, Barriers to Energy Efficiency. In our community, there are almost as many recommendations available for data center professionals as there are data center energy consultants. Still, however, offered advice is shelved for future consideration – saved, perhaps, for a time when money or resources are available for side projects. For those of us who believe that energy efficiency pays for itself – either directly in operational expenses or indirectly through deferred capital expenditures – resolving these barriers is the crux of the problem.

One particular barrier lies in the split incentives that exist between IT and facilities organisations. Although this is changing in a few places, situations where IT is not responsible for paying the energy bill are common, if not typical. The latest thinking, however, suggests that not only is IT responsible for the key policies that have the greatest effect on data center energy consumption, but they are also directly affected by these decisions. Smart deployment and operational decisions both reduce data center energy consumption and significantly improve data center capacity and lifetime.

So, it’s time to refresh – not only our servers but also our plans, policies and thinking. With an update to the data and analysis that are the foundation of our current efforts, I believe we will find that we are on the road to recovery. The adoption of virtualisation, the development of productivity-related metrics, and the continued improvement in server performance per watt all point to a bright future for the energy efficient data center. The next question, however, is how can we keep up with the ever-increasing demand for computation while keeping energy consumption as low as possible. That is a subject for another day.

This article first appearedin DatacenterDynamics FOCUS

Keywords: Dell, IT, power, data center, efficiency, energy, ocnsumption