Transport-Layer Control to Increase Throughput in Bufferless Optical Packet-Switching Networks

We show that, in an optical packet-switching (OPS) network with bufferless switches, end-to-end congestion control algorithms can be designed to completely overcome the pernicious effects that the lack of buffers have on application throughput. Thus, applications can obtain the same throughputs in bufferless OPS as in equivalent electronic networks with plentiful buffers. There is no need to implement buffer emulation techniques such as fiber delay lines or to use resource reservation or distributed scheduling such as in optical circuit switching and similar techniques. Critically, this implies that many OPS switch prototypes and testbeds developed by the research community can be used to realistically compete with electronic packet switching (EPS) regarding application throughput without the need to wait for any kind of buffering technology. Accordingly, we present two congestion control algorithms to showcase this in the context of TCP flows: stop-and-wait, which is focused on local area networks (LANs), and modified additive-increase-multiplicative decrease, which is focused on large LANs and metropolitan area networks.