Optical versus electronic packet switching in delay-sensitive 5G networks: myths versus advantages

We show that optical packet switching (OPS) 5G networks without buffers in their switches are expected to offer applications that require reliable packet delivery [e.g., zero losses via transmission control protocol (TCP)] the following improvements, compared to what is offered by the state-of-the-art in electronic packet switching (EPS) networks: lower flow completion times, and thus higher throughputs (useful for most Internet applications); lower probability ofmissing hard per-packet end-to-end deadlines (useful for real-time applications); and higher numbers of packets that outrun their counterparts that use an EPS network (useful for winner-takes-it-all applications like latency-sensitive trading algorithms).We also show that this superior performance of OPS over EPS increases asymptotically in favor of OPS as network loads grow. We show how this counterintuitive result seems to stemfromthe lowaverage and low variance of end-to-end packet delays caused by the lack of queues in bufferlessOPS switches,which, in turn, allow end-to-end congestion control algorithms (e.g., inside TCP) to react and compensate packet losses swiftly.