A Novel Temporal Dynamic Wavelength Bandwidth Allocation Based on Long-Short-Term-Memory in NG-EPON

Abstract

Optical networks have undergone a remarkable transformation with the adoption of Artificial Intelligence (AI) techniques such as Machine Learning (ML) and Deep Learning (DL). The Next Generation (NG)-EPON is one such technology that is essential for supporting high-bandwidth applications like 4K video streaming, ultra-high-definition (UHD) CCTV, and other emerging video-type applications that have strict Quality-of-Service (QoS) requirements. In this paper, we present a ground-breaking Temporal Dynamic Wavelength Bandwidth Allocation (T-DWBA) mechanism based on the Long-Short-Term-Memory (LSTM) architecture. The T-DWBA uses past experiences to learn data as knowledge and predict time series with time lags of unknown size. Our proposed mechanism reduces upstream control message overheads, eliminates idle periods, and significantly improves bandwidth utilization, ensuring superior QoS specifically for video-type applications. The simulation results show that the T-DWBA significantly enhances system performance, reducing packet delay, and jitter, and improving bandwidth utilization. The use of AI techniques like ML and DL coupled with the recent advancements in SDN-Enabled Broadband Access (SEBA), hardware/software, and cloud technologies provide the perfect platform for deploying our proposed T-DWBA mechanism. Overall, our research proposes a promising solution for boosting EPON performance, revolutionizing optical networks, and providing seamless access to high-quality video streaming for a next-generation audience.

Keywords

• NG-EPONs
• T-DWBA
• LSTM
• SEBA
• system performance