An Adaptive Model of Energy Consumption Predictor for Big Data Centers

Abstract

It is well-known that with the explosive growth of data, the age of big data has arrived. However, the power consumption of the big data center is huge and will be a major obstacle to its wider extension. How to save the power towards green computing is a potential tendency by utilizing the elastic computing capability to analysis the power consumption over cloud computing platforms in big data centers. Anyway, the high energy cost of data centers have highlighted the energy models in the configuration of cloud providers' big data centers, which can estimate the energy consumption and adopt specific strategies to save the power within a given power budget. Moreover, the design of an adaptive energy model which can self-model the energy consumption on different conditions is quite challenging research. In this paper, we propose pBigData (Power Saving for Big Data Centers), an adaptive self-modeling paradigm, which can predict data centers' energy consumption, warn to exceed the threshold value and dynamically construct new model when existing ones become inadequate due to changes in hardware or different workloads. pBigData is to monitor the energy consumption according to the collecting data of devices including disks, processors, networks and temperatures, and then provide in-depth statistical analysis of the energy consumption. Our experiments reveal conclusively how accurate pBigData can enhance system energy efficiency while maintaining performance. Index Terms - Big Data, Data Centers, Power Saving, Energy Consumption, Predictor, Early-warning