Architectural options for the next-generation networking paradigm: Is Optical Internet the answer?

Recent advances in optical networking technologies are setting the foundati on for the next-generation data-centric networking paradigm, an "Optical In ternet". This work addresses one of the most challenging issues facing toda y's service providers and data vendors; how will the SONET/SDH-based legacy infrastructure currently in place make a graceful transition to the next g eneration networking paradigm? A simplified, two-tiered architecture that requires two types of sub-system s will set the stage for a truly optical internet: service delivery platfor ms that enforce service policies; and transport platforms that intelligentl y deliver the necessary bandwidth to these service platforms. If IP can be mapped directly onto the WDM layer, some of the unnecessary network layers can be eliminated, opening up new possibilities for the potential of collap sing today's vertically layered network architecture into a horizontal mode l where all network elements work as peers to dynamically establish optical paths through the network. This paper presents a balanced view of the visi on of the next-generation optical internet. The work presented here builds on the IETF multi-protocol lambda switching (MP lambdaS) initiative and addresses the implementation issues of the path selection component of the traffic-engineering problem in a hybrid IP-cent ric DWDM-based optical network. An overview of the methodologies and associ ated algorithms for dynamic lightpath computation is presented. Specificall y, we show how the complex problem of real-time provisioning of optical cha nnels can be simplified by using a simple dynamic constraint-based routing and wavelength assignment (RWA) algorithm that compute solutions to three s ub-problems: (a) routing; (b) constrained-based shortest-route selection; ( c) wavelength assignment. We present two different schemes for dynamic provisioning of the optical ch annels. The two schemes use the same dynamic lightpath computation approach proposed here, except that the third component, that is the wavelength ass ignment algorithm and its implementation, is different for each of the two schemes.