In our last update we talked about Power Usage Effectiveness (PUE). The calculation measures how effective your data center is in using the input power. This ratio of power available to power used will yield a factor greater than 1. The larger the number the less efficient your utilization is.
More than IT
As technologists our first reaction is it to look at what we control – servers, network and storage. Part of any strategic architecture discussion you have with your partners and vendors needs to include the operating efficiency of their product. Our world is becoming dense. Virtualization and the normal progression of processing power is driving more function into a tighter space. Efficiencies are gained in space at the cost of kilowatts per square foot.
As our IT world changes, so does the world of the ancillary products we draw on to keep the servers, network and storage working. Cooling units, power distribution units, UPS systems and other elements in the data center are part of the equation. Every year unveils improvements in those components which lead to less energy need to cool the same space or more effective distribution of power. More efficiency in those components can change the PUE.
Decisions to make those kind of infrastructure upgrades can be difficult. As an example, the air conditioning units or CRUs (computer room units) might be on the same depreciation schedule as the facility. A decade ago having flexibility to upgrade CRUs was not a high priority since the cost of the upgrade outweighed the benefits of more efficient units. With the ongoing rise in energy costs, coupled with greater efficiency in modern CRU’s, that decision point may be changing.
I recommend that you have a straightforward conversation with your finance and facilities teams about how best to gain flexibility going forward. It might be changing depreciation schedules for select components, or an agreement on a set of criteria that would trigger the review of an upgrade, at which time the writeoff of the existing units would be part of the evalution.
The Build or Buy Impact
Understanding PUE can also help you decide to build or buy. The data center is a dynamic environment. Over time there will be more demand for power and cooling as density increases. More functions, like spam filtering, optimization (ADO) and load balancers have taken up rack space. They are the beginning. TCP/IP based voice systems, which are server based, and mobile application stores need to be housed somewhere. The demand for space, power and cooling will increase, but not only from traditional components.
The picture you build for your internal data center must be based on what it will need to support 5-10 years from now. If you do not have a good sense based on a technology roadmap, then choose a planning number, such as 10% per year growth, for incremental demand as a projection. That provides you a baseline to understand what equipment your data center needs to sustain your growth. It means your PUE might be worse now than in the future.
Knowing what it will cost you for IT infrastructure and the support infrastructure will allow you to assess the benefit or cost of a colocation strategy. It really is a matter of risk. Do you want to take on the risk for planning power and cooling capacity as well as computing capacity? There are many factors that go into a build versus buy decision and one of them should be the relationship between the IT infrastructure and the infrastructure that gives it the power and cooling it needs. PUE won’t give you the answer, but it will provide you some insight to make a more informed decision.
To get more useful data center management strategies from Nemertes Research download the Q1 2012 Data Center Knowledge Guide to Enterprise Data Centers.