Dimensioning WDM Networks for Dynamic Routing of Evolving Traffic

New Internet applications are increasingly generating high-bandwidth and short-lived demands. If network resources are available, establishing a lightpath on demand only takes a few minutes on today's reconfigurable optical networks. These demands thus create a more variable and unpredictable environment for long-term network planning. At the same time, upgrading backbone networks remains expensive and infrequent. Dimensioning network resources to sustain variable traffic demands for a long time, while requiring fewer upgrades to achieve high performance, has become a challenging problem. Two kinds of dimensioning problems for optical opaque networks are proposed and studied in this paper: basic dimensioning allocates network resources for a newly built network, and incremental dimensioning allocates extra resources for future demand growth and variations. We propose new metrics to quantify the traffic load and the traffic pattern evolution for dynamically routed networks. We evaluate the performance under load scaling, traffic evolution, and misdimensioning. We show that a dimensioned network can sustain a much higher load while providing the same performance compared with misdimensioned ones. Our approach is better adapted to traffic evolution than a uniform allocation and an asymptotic optimization approach proposed earlier.