Abstract:
The extensive growth in the demand for network bandwidth has created new challenges for net- work operators in the last few years. Traditional optical network technology is no longer enough to handle the massive bandwidth requirements. Hence to meet the ever-increasing demand for bandwidth e ectively, the Elastic Optical Network (EON) paradigm has been designed. With dynamic tra c in a network, the spectrum becomes fragmented due to tra c coming in the net- work at di erent times with varying holding time. This fragmentation within the spectrum leads to the blocking of future demands due to the non-ful llment of optical constraints. Therefore, defragmentation is performed to improve spectrum continuity and e ciency. However, re-tuning any connection during defragmentation a ects the physical properties of other connections. Furthermore, the repeated computation of certain physical properties of the network may be- come computationally expensive over time. This report proposes a heuristic algorithm to miti- gate fragmentation for the C+L band network scenario in static and dynamic tra c optimally. The report also proposes a robust Neural Network learning framework to predict the OSNR and accelerate the defragmentation process.
