LSTM-based Traffic Prediction in Support of Periodically Light Path Reconfiguration in Hybrid Data Center Network
Haoxiang Shi, Cen Wang · 2018
Hybrid network combined with packet switching and optical switching has become a compromise in modern data center to provide high capacity yet flexibility. The light path reconfiguration in such network is essential to take advantage of optical switching. However, the light path reconfiguration and the traffic variation may mismatch because that the control procedure may not catch up with the dynamic traffic, which will deteriorate the network performance. Therefore, in this paper, we propose a machine learning-based, specifically the LSTM-based traffic prediction method to assist light path reconfiguration in hybrid data center networks (DCN). Simulation and experimental results verify its effectiveness of exploiting network agility to promote cloud computations with lower job completion time.