AWGshare: A highly scalable and strictly non-blocking optical switch

This article presents a strictly non-blocking switch design for all-optical nodes. It is particularly suited for core nodes supporting hundred(s) of wavelengths (W) per fiber. The first part presents several scalability issues involved with existing designs and justifies the choice of switch architecture and optical components. The second part presents the three stage AWGshare, whose switching elements are based on arrayed waveguide gratings (AWG) and tunable lasers. It proves to have the same connectivity as the classic Clos architecture, but differs in implementation by using a single switching element at the second stage, thus saving 2W-2 AWGs. We also show that implementing B fully shared buffer-interfaces, whilst keeping the strictly non-blocking property of the switch, requires at least W-1 more buffer-interfaces for the classic Clos implementation than for the AWGshare. The last part compares the AWGshare and six other strictly non-blocking designs from literature with respect to component size, number and physical characteristics. The AWGshare shows a good balance of the mentioned properties. When also accounting for the low cost of implementing fully shared buffers, it stands out as an attractive switch design for future all-optical core nodes.