Zeitschrift für Informatik und Systembiologie

Several surveys of the findings of research into the power consumption of virtual entities (VEs, also known as virtual machines (VMs) or containers) have been published. Our contribution to this work is a thorough examination of the dynamics of research itself: the challenges, approaches, pitfalls, fallacies, and research gaps, without ignoring the research results. The prospective researcher who wants to understand the dynamics of research into predictive modelling and supporting measurements of power consumption by individual VEs relevant to the telco cloud is our target audience. A thorough frequency analysis is used to characterise dynamics, which we do using a novel method we developed that is unique in its ability to parse research literature. A prospective researcher can obtain a thorough characterization of the problems, approaches, developments, formal methods, pitfalls, fallacies, and research gaps that characterise this research space by using the visual aids we provide and our observations through cross-cutting themes. Among the themes identified by our survey, we noted that all of the problem categories we identified touch on one or more of a set of seven major variables that may affect power consumption by virtual entities and the resulting model representations: workload type, virtualization agent (VM or container) characteristics, host machine resources and architecture, temperature, operating frequency, attribution of a fraction of consumed power.