Dynamic Bandwidth Defragmentation in TWDM-PON Based Ethernet Fronthaul

Using Ethernet in fronthaul can achieve statistical multiplexing gains based on the new function split. When TWDM-PON is used in Ethernet based fronthaul network, the large granularity of bandwidth requirements and the diversified service types will lead to the bandwidth fragmentation in OLT pooling. To solve this challenge, we develop a mathematical model by establishing the 0-1 relationship matrix among ONUs, OLTs and wavelengths. In the proposed location-based adaptive genetic algorithm, we capture the nearly optimal relationship matrix, improving the performance of network resource allocation. Under the assumed conditions, the simulations show that the proposed dynamic bandwidth defragmentation scheme effectively reduces the degree of bandwidth fragmentation from 0.7 to less than 0.2 under the limited strength of the migration.