Allocating Data Center Energy Costs and Carbon to IT Users

by Neil Rasmussen

Executive Summary

Are complicated software and instrumentation needed to measure and allocate energy costs and carbon to IT users? Or can we get by with simple, low cost methods for energy cost and carbon allocation? How precise do we need to be?

This paper provides an overview of energy cost and carbon allocation strategies and their precision. We show that it is both easy and inexpensive for any data center, large or small, new or old, to get started allocating costs and carbon, but the expense and complexity escalate and ROI declines when excessive precision is specified.

Conclusion

This paper has described a logical strategy for allocation of data center energy and carbon to IT users.

Simple, no-cost models of energy use can be used to allocate energy and carbon based on standard, average units of IT capacity, such as a standard “server unit”. Such models are not precise, but have sufficient accuracy to be useful in a data center energy management system.

A simple system can be enhanced over time by adding additional measurement and modeling capabilities, providing additional precision and insight into energy use. This paper outlines a rational sequence of implementing these capabilities. A surprisingly effective and low-cost system can be implemented with only a small number of energy meters combined with a specialized data center energy audit and simple software.

Data center operators should not assume that complex and extensive metering systems are needed in order to implement an effective energy management system, or to assign energy and carbon to IT users. In fact, extreme instrumentation systems have a poor return on investment.

This paper outlines a starting approach that any data center operator, large or small, can implement immediately at no cost. Every unnecessary watt of power drawn by a data center represents an unrecoverable loss. A crude but simple energy management system implemented today is much more effective than an ideal system implemented later, because no matter how effective an energy management system is, it can’t recover energy already lost. 

Download PDF