Network re-optimization is a process that must be triggered periodically in order to improve the nefficient iresource allocation of online routing heuristics due to the uncertainty of online lightpath demand arrivals and departures. Network re-optimization involves two tasks: a) finding new lightpaths for a (sub)set of current demands, i.e. rerouting, and b) migrating the current traffic to the new configuration diminishing traffic disruptions, i.e. lightpath reconfiguration. If not controlled, excessive traffic disruptions may be a cause of violations of clients’ Service Level Agreement, which should be compensated by the network operator with penalization fees.
So far, rerouting and reconfiguration tasks of a re-optimization process have been done separately, hence, efforts in trying to achieve the best network performance (rerouting) yields to solutions incurring on unacceptable traffic disruptions (reconfiguration) and vice-versa. In this paper, given a time-disruption threshold for reconfiguring every demand, we present a novel methodology consisting on two procedures that collaboratively find the best network performance without incurring on penalization fees.
Our numerical results are extremely encouraging: in our scenarios, it is always possible to achieve an optimal routing performance without incurring on penalization fees.