Bachelor Thesis: Evaluating the Relations between Energy Efficiency Metrics

Motivation

With the regular release of ever more powerful hardware and the rising size of data centers, energy consumption and efficiency of servers has become a growing concern. Governments, academic institutions, and industrial manufacturers seek ways to improve and measure the energy efficiency of these devices. As a result, new metrics for the measurement of servers’ energy efficiency are being developed. Additionally, existing metrics from other fields of computer science, such as battery efficiency of mobile systems, are being appropriated for use in a data center context. Relations between these metrics are not thoroughly explored, as is their expressiveness. 

 

Goals

The focus of this thesis will be a summarization of existing energy efficiency metrics and their relations to one another. For that purpose the planned activities are:

1. Research on existing energy efficiency metrics in the field of data center energy measurement and beyond. Metrics will be categorized based on the dimensions contained within them. For metrics from other fields of computer science, their applicability in the data center context will be evaluated.
2. Creation of a data center power simulation for the evaluation of metric expressiveness. Metrics will be tested against different configurations in server workload and hardware. Each metrics’ reaction to these changes will be evaluated.

 

We offer

• Collaboration with leading industry server providers, such as HP, Fujitsu
• Results will be fed into benchmark standardization efforts at SPEC
• Excellent working environment and intensive mentoring
• Work with state-of-the-art and innovative technologies

Duration

3 months

 

Contact

Jóakim v. Kistowski, M.Sc.

joakim.kistowski@uni-wuerzburg.de