View article

The growing appetite for Cloud services leads to an unprecedented increase in the electricity consumption of Data Centers. For instance, according to a recent report, in 2014, data centers in the US consumed about 1.8% of total US electricity [5]. This energy consumption is among the sources of pollution and global greenhouse gas emissions of Information and Communication Technologies (ICT), with the extraction of raw materials and end-of-life product recycling. To reduce the negative impacts of ICT on global warming, major Cloud actors like Amazon AWS, Apple or Microsoft are involved in projects to deploy solar power facilities [4]. This project is a follow-up of papers [1, 2, 3]. It focuses on a geographically distributed cloud. Virtual machines (VM) arrivals are considered unpredictable and no assumption will be made concerning future submissions. Each data center (DC) is supplied by a farm of photovoltaic panels (PV). The PV forecast model is based on a truncated normal law. For each data center and at each future time slot, an expected value of generation and a variance value compose the truncated normal law forecasting the PV generation. One challenge is to formulate a realistic communication network between data centers, including energy consumption. This energy consumption cannot be considered as an on-site consumption for the sending or receiving data center, and is thus not concerned by local PV production. The first objective consists in developing a realistic model of Clouds supplied by green energy, and a simulation platform to compare scheduling algorithms. This model and this platform will then be used to examine …