LSTM for Cloud Data Centers Resource Allocation in Software-Defined Optical Networks

Abstract

Nowadays, artificial intelligence provides an excellent opportunity for scientists to improve the efficiency of resource allocation in communication networks. In this paper, we focus on applying two methods: Long-Short Term Memory and Monte Carlo Tree Search, to solve the problem of cloud resource allocation in dynamic, real-time traffic scenarios. We use a framework of Software Defined Elastic Optical Networks and cloud resources available from Amazon Web Services. Results show that the application of Monte Carlo Tree Search and Long-Short Term Memory provides superior performance, which is an excellent opportunity for network operators to achieve better utilization of their networks, with lower operational costs.