The Efficiency Frontier: Optimal Combinations of Energy-Saving Strategies in Data Centers

With the growth of electricity demand for data centers projected to grow at an annual rate of 27% annually, greater attention is being paid to the impact on the grid, environment and communities. Traditional approaches to im-prove power utilization efficiency (PUE) in data centers often evaluate strategies individually. This involves either a simple “go/no-go” decision for a given strategy or selecting the single “best” strategy from a set of candidates. This research posits that considering combinations of strategies expands the solution space beyond individual strategies, increasing the likelihood of finding a superior solution. By drawing on principles of system architecture and multi-domain design optimization, this challenge is framed as a resource optimization problem, and a linear program is formulated to solve it. The value of this framework is demonstrated by revisiting a previous study that examined energy efficiency strategies in isolation and stacked them over each other. The combination of two and three strategies resulted in an improvement in PUE of 18% and 7%, respectively, compared to the cumulative PUE achieved by evaluating these strategies individually and stacking them over each other. The results suggest that the impact of this framework is highest when the number of strategies is two, and decreases as the number of strategies increases.