Economic framework for resource management in data centers

Data centers consume more energy than ever. However, recent works prove that one third of the total energy consumed is wasted by keeping servers in comatose and almost half by keeping them in comatose and idle. To address this issue, we have developed an economic framework and a model based on the servers On-Off technique and semi-Markov Decision Process. This approach provides an optimal policy that minimizes data center's power consumption and maximizes the Cloud Provider's profit. The model allows updating dynamically the number of active servers in a data center according to the traffic. In the tested scenarios the proposed approach saves up to 66% of total energy consumed.