1. Introduction
Use of optical interconnects in data center networks is a key requirement to achieve Internet-scale data centers due to the potential and benefits of optical components [1]. Optical interconnects are power efficient and support huge bandwidths. Current interconnect designs can be categorized into four types based on the optical technology used: microelectromechanical system (MEMS) optical cross connects (OXCs), arrayed waveguide grating routers (AWGRs), semiconductor optical amplifiers (SOAs), and wavelength selective switches (WSSs). MEMS switches are based on mature technology, have small insertion loss and cross-talk, and are data rate independent. They are inexpensive and are highly scalable. However, the switching time of these switches is in the order of tens of milliseconds [2], [3]. An AWGR switch is a passive device and works in combination with tunable wavelength converters (TWCs) or tunable lasers (TLs) while an SOA works as a gate element that manipulates light and also compensates for losses that occur during transmission of optical signals. A large switching fabric using AWGR switches has been achieved by cascading AWGRs in a three-stage architecture [4]. Similarly, a large switching fabric using SOA-switches can be obtained by arranging SOAs in a broadcast-and-select architecture [5] or in a three-stage Clos architecture [6]. AWGR and SOA switches have switching times in the range of few nanoseconds [5] – [7] but they are expensive in comparison to MEMS switches of the same capacity. A WSS divides incoming set of wavelengths to different outgoing set of wavelengths. Each set of wavelengths is destined to a specific output port. The switching time of a WSS is in the order of tens of microseconds [8].