A Learning Living Network With Open ROADMs

We propose a living network as a key enabler for open reconfigurable optical add/drop multiplexer (ROADM) networks that, unlike today's use of static planning tools, adapts to varying network conditions, and allows network operation close to actual performance. We experimentally demonstrate the living network in various network conditions that autonomously keeps record of its path-level performance. The more services are added, the more accurate the performance of a newly to be established service is estimated through a learning process, which shows the feasibility of living network with an allocatedmargin of only 1.2 dB.