Data centers are essentially living, breathing entities that change over time. The problem is, it’s difficult to predict in advance how they’re going to change – and therefore to anticipate power requirements.
Indeed, when many companies build data centers they purposefully leave room for future growth, as they should. When doing so, however, it’s not always a good idea to specify power density requirements in advance for rows or zones that are not yet planned, simply because your future requirements are difficult to predict with any accuracy.
It may be tempting to build out all the power and cooling infrastructure in advance to support a pre-defined power density. The advantage of this approach is that when the time comes to deploy new IT infrastructure, you don’t need to be performing major utility work in a live, production data center.
However, this approach has a number of major potential downside costs, including:
- Future IT density exceeds the density of the power and cooling infrastructure and therefore cannot be effectively deployed
- Future IT density is less than the density of the power and cooling infrastructure and therefore major infrastructure investments are wasted
- The facility never expands, or expansion is forced to occur in another location due to unforeseen circumstances, so infrastructure investments are wasted
- The near-term loading on the data center is much lower than the rated power and cooling infrastructure, leading to major decreases in electrical efficiency and inflated electricity costs
- The advance installation of currently unneeded power and cooling infrastructure drives unnecessary capital equipment and maintenance contract costs
To avoid these issues requires employing a design and implementation approach that uses modular, scalable power and cooling infrastructure. Such an architecture would be based on the up-front installation of main utility feeds, such as row or zone-level power feeds and cooling feeds.
In this way, decisions on the specific density to be supported within a zone or row can be deferred to the time of deployment, and the power and cooling infrastructure would be deployed on a row-by-row basis. The APC by Schneider Electric InfraStruXuresystem is one practical example of such architecture.
One key element of this approach is to plan for the worst-case power density value in any rows or zones that are to be deployed in the future, and to install primary feed wiring and piping up front to support that density. However, you defer the selection and deployment of power and cooling equipment until you know what the actual power density requirements will be.
The benefit of this approach is you’re able to right-size the power and cooling infrastructure to the actual application, and deploy it only where and when it’s needed. This greatly reduces capital costs, operating costs, and results in a much more energy efficient data center.
To learn more about how to properly plan for data center power capacity, read the
APC by Schneider Electric white paper, “Guidelines for Specification of Data Center Power Density.”