Thermal Prediction for Efficient Energy Management of Clouds using\n Machine Learning

Thermal management in the hyper-scale cloud data centers is a critical\nproblem. Increased host temperature creates hotspots which significantly\nincreases cooling cost and affects reliability. Accurate prediction of host\ntemperature is crucial for managing the resources effectively. Temperature\nestimation is a non-trivial problem due to thermal variations in the data\ncenter. Existing solutions for temperature estimation are inefficient due to\ntheir computational complexity and lack of accurate prediction. However,\ndata-driven machine learning methods for temperature prediction is a promising\napproach. In this regard, we collect and study data from a private cloud and\nshow the presence of thermal variations. We investigate several machine\nlearning models to accurately predict the host temperature. Specifically, we\npropose a gradient boosting machine learning model for temperature prediction.\nThe experiment results show that our model accurately predicts the temperature\nwith the average RMSE value of 0.05 or an average prediction error of 2.38\ndegree Celsius, which is 6 degree Celsius less as compared to an existing\ntheoretical model. In addition, we propose a dynamic scheduling algorithm to\nminimize the peak temperature of hosts. The results show that our algorithm\nreduces the peak temperature by 6.5 degree Celsius and consumes 34.5% less\nenergy as compared to the baseline algorithm.\n