Power Consumption in Bufferless Optical Packet Switches in SOA Technology

Increase in data transmission and processing speed unavoidably leads to high requirements on power supply. Especially in the case of high-capacity electronic routers, the question of power consumption will be the major issue and probably the most important limiting factor for the future. This paper analyzes the role of bufferless optical packet switches (OPSs) realized in semiconductor optical amplifier (SOA) technology in reducing average power consumption. An analytical model is proposed to provide an expression for the average power consumption of bufferless OPSs as a function of both the offered traffic and the main parameters characterizing the power consumption of the space switching modules (SSMs) and wavelength converters (WCs) needed to realize the bufferless OPSs. The effectiveness of WC sharing, which reduces the average power consumed by WCs, is evaluated in the case in which the Tucker’s model for the evaluation of the power consumed by the SSMs and WCs is used. The obtained results show that, due to the high power consumed by the switching elements, switches with shared WCs may require average power consumption higher than switches in which WC sharing is not performed. In particular, only when the offered traffic is low and high-energy-consumption WCs are used, is the WCs sharing effective. When cross-gain modulation WCs are employed, the sharing technique allows for a reduction of average power consumption from 50% to 10% when the offered traffic is varying from 0.1 to 0.9. To obtain more realistic power consumption of the bufferless OPSs, we have introduced in our analytical model some power consumption values evaluated and experimentally validated for the SOAs produced by some manufacture’s and needed to realize the SSMs and WCs. In this case we compare the power consumed of the bufferless OPSs to that of a Cisco GSR 12008 router equipped with ten slots that can accommodate up to $4\mathrm{Gb}∕\mathrm{s}$ per slot and designed for operation in a network core. We take into account for comparison the power consumption normalized to the offered total input bit rate.