Almost-optimal design for optical networks with Hadoop cloud computing: Ten ordinary desktops solve 500-node, 1000-link, and 4000-request RWA problem within three hours (invited)

It was found that the sequence of lightpath demand list plays an important role for the performance of routing and wavelength assignment (RWA) algorithm [1]. We develop a ten-desktop Hadoop cloud computing system with each desktop independently running the RWA algorithm for a certain number of demand sequences such that a sufficient number of demand sequences can be evaluated within a short time. We compare the results of all the evaluated demand sequences to choose the best one as the final solution to the RWA problem. Simulation studies show that the approach of evaluating multiple shuffled demand sequences can achieve performance same as (or very close to) the optimum. For a large network with 500 nodes, 1000 links, and 4000 requests and 5000 parallel shuffled lightpath demand sequences, we demonstrate as a record that the Hadoop system is efficient to run the same RWA algorithm for all the sequences within 3 hours, 30 times faster than a single ordinary desktop.