Pitfalls in energy consumption evaluation studies

Energy consumption is fast emerging as a central item of research for the future Internet (FI). In an environment where the majority of users access next generation networks using battery-powered devices, energy efficiency is of equal importance to other well-established metrics of overall system and protocol performance. Nevertheless, we still lack benchmarks of what constitutes energy efficient operation in mobile computing and networking. In particular, models that accurately quantify transport protocol energy consumption are in short supply. This paper introduces a blueprint for a configurable and extensible energy consumption model which takes battery depletion and recovery characteristics into consideration. We show that the non-linear nature of battery depletion cannot be ignored in energy consumption evaluation studies. Battery recovery effects should be accounted for as well. We explore the potential gains in different TCP variants using our model and illustrate that data-counting and time-counting are not sufficient methods for assessing the energy efficiency of transport protocols operating on devices with fixed-energy sources.