Optical Data Center Networking: A Comprehensive Review on Traffic, Switching, Bandwidth Allocation, and Challenges
Peristera Baziana
Abstract
The accelerated growth of data traffic in data centers (DCs) globally is driven by the dominance of multiple emerging data-intensive applications hosted by edge/cloud DCs, such as Internet of Things (IoT), augmented and virtual reality etc. The advent of machine learning/artificial intelligence (ML/AI) and its exploitation in several applications for prediction and optimization purposes has further burdened the traffic conditions within existing DC networks (DCNs) where traffic burstiness and high variability dominates. Most of the existing DCNs worldwide use conventional electrical switches which cannot effectively serve the highly differentiated traffic, while optical switching is currently used by few big-data companies’ DCs and only in the core and aggregation DCN tiers. As a result, limited DCN performance in terms of end-to-end delay and bandwidth utilization is provided. Therefore, the investigation on all aspects of optical networking (switching, transmission techniques etc.) in DCs attracts high attention in both academia and industry, aiming to provide high DCN bandwidth utilization, low end-to-end latency (in the order of few <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\mu $ </tex-math></inline-formula> s), and low data loss. In this paper, a comprehensive review on traffic characteristics, switching techniques, bandwidth allocation schemes and power consumption reduction technologies for next-generation optical DCNs is extensively given. Moreover, trends of optical technology enabled in optical DCNs are comprehensively discussed, putting in the agenda the technical challenges and limitations towards efficient optical DCNs capable of serving the stringent traffic classes demands for guaranteed low completion time of diverse edge/cloud data-intensive applications.