Extra Buffer Resources Improving Competitiveness in Minimizing Energy Consumption

In this paper, we consider energy management algorithms for scheduling jobs in power-scare scenarios such as embedded computer systems and sensor networks. We focus on investigating the impact of buffer resources in minimizing the total energy cost in an online setting. The online algorithms do not have any assumptions on job arrivals; their worst-case performance is measured in term of competitive ratio, when they are compared with the optimal algorithms with clairvoyance. We prove that with appropriate extra buffer space, an online algorithm can beat an weak optimal offline algorithm in terms of the total energy required. Our research result helps to quantitatively estimate the optimal on-chip buffer resources allocated in real-time systems with power constraints. We also present the lower bound of competitive ratio that any deterministic online algorithm cannot achieve.