Reducing the Operational Costs of Data Centers through Deep Q-Learning

  • Abhijeet Nawale, Rajeshkannan Regunathan


Deep Q-learning is the incorporation of Q-Learning and Artificial Neural Network. Q-Learning is a reinforcement learning technique, which uses rewards to achieve the goal. This paper discusses about developing an AI such that it maintains the temperature of a server in an optimal range. Server environment will be set up, and the AI will be controlling the cooling/heating of the server so that it stays in an optimal range of temperatures while saving the maximum energy, therefore minimizing the costs using deep Q-learning. The number of users and the rate of data transmission will be randomly fluctuating to simulate an actual server. This will lead to randomness in the temperature and the AI has to understand how much cooling or heating power it has to transfer to the server so as not to deteriorate the server performance and, at the same time, expend the least energy by optimizing its heat transfer.

 Keywords: Reinforcement learning, Deep Q-Learning, Activation function, Optimizer, Neural network, Sigmoid, Hard sigmoid, Rectified Linear Unit, Stochastic Gradient Descent, Adaptive Moment Estimation, Adaptive Gradients.


How to Cite
Abhijeet Nawale, Rajeshkannan Regunathan. (2020). Reducing the Operational Costs of Data Centers through Deep Q-Learning. International Journal of Advanced Science and Technology, 29(05), 2949 - 2966. Retrieved from