Operating a Network Close to the 'Zero Margin' Regime Thanks to Elastic Devices

Despite of recent progress in research on the subject of modernisation and automation of networks, most legacy optical networks still operate statically: when a connection has been set-up it is rarely reconfigured. Exception occurs in a case of a node/link failure: in that case reconfigurations concern only the traffic with a specific class of service (CoS) where a given resiliency has to be ensured. When network reconfigurations are needed, the newly set up connection has to be promptly operational, independently of the actual network state. To ensure the fast set up of a connection, it is mandatory to have physical estimators defining the feasibility of a connection depending on the state of the network: the capacity and placement of neighboring channels, the degradations of the network components due to aging, statistical penalties, etc... In order to guarantee the operational set up of a connection all the time, today various margins are introduced to ensure reliable network operation [1]. Firstly, system margins include transponder and fibre ageing as well as polarization dependent loss (PDL) and non-linear impairments (NLI); secondly, design margins are due to the uncertain knowledge of the physical characteristics of the system devices and predictor tool inaccuracies; and finally, unallocated margins as a by-product of network traffic demand heterogeneity in reach and capacity associated with a chosen data rate [2]. All these margins in total can attain 6dB [3]. In the future, due to cloud computing, high data rate multimedia and the higher distribution of data centers, increasingly large data transfers are forecast placing new requirements and challenges on backbone networks [4] relative to their dynamic nature. Because of this dynamic nature, it is possible to avoid some margins (system and unallocated margins, called operation margins in the following), by collecting information on the physical state of networks using real time optical performance monitoring. Thanks to coherent receivers, much information on the state of the network can be provided directly by the transponder, and no longer by special monitoring devices [5]. Moreover, thanks to digital signal processing (DSP) enabled by advanced programmable digital-to-analog convertors (DAC), a transponder can dynamically adapt its capacity, adapting its modulation format/symbol rate/code correction as a function of the physical degradation on a given path [6].