Algorithms for Scheduled Lightpath Switching in Optical Networks with Channel Impairments

Applications

The Inventors have developed a lightpath scheduling algorithm to facilitate optical communication via Erbium-doped fiber amplifier (EDFA)-based optical networks. 

Problem Addressed

Present-day networks are being challenged by dramatic increases in data rate demands of emerging applications. Incorporating fast lightpath switching into optical network architectures is critical to enable the growth, power efficiency, and cost-effective scalability of next-generation networks. A lightpath is an end-to-end all optical connection from the source node to the destination node composed of a constant wavelength. Scheduled lightpath switches improve network resource utilization while maintaining a good network performance by relinquishing network resources to other users. The Inventors’ algorithm for lightpath switching utilizes models of physical layer impairments to make sure the scheduled lightpath is not corrupted by switching induced channel impairments in EDFA-amplified optical networks

Technology

This patented technology predicts the channel quality of a lightpath based on how EDFAs are distributed along the lightpath and the number of channels that are lit up in the same fiber. The scheduling algorithm chooses the first available lightpath that can accommodate the requested transmission duration and does not overlap in time with other scheduled sessions on the same wavelength along a candidate path. The first available lightpath is then checked for quality using the models at times during the transmission session when there is a change of channel configuration (i.e. the number of channels that are lit up in a fiber). If the lightpath is not qualified at one or more times of change of channel configuration, the algorithm recursively starts the searching of the next first available lightpath starting from the time when the lightpath fails the quality check. 

Advantages

  • Easy-to-implement algorithm that needs far fewer steps and less time to light up one wavelength (~50ms) than minutes for existing techniques
  • Predictive models of channel quality can be used to greatly reduce the amount of network sensing, reporting traffic and control plane complexity