Online Resource Allocation in Dynamic Optical Networks

Abstract

This paper presents a synopsis of ongoing research in the field of resource allocation in dynamic optical networks. This type of networks are envisioned to serve connections with random arrival and holding times and with fast connection setup requirements. Based on this, we model resource allocation as a mechanism that includes network control functions for routing and spectrum assignment (RSA), connection admission control (CAC) and grade of service (GoS) control. The goal is to efficiently assign spectrum resources to connections so as to attain optimum performance. For that, the network is modelled as a stochastic loss system subject to spectrum continuity and contiguity constraints. The theory of Markov decision processes (MDP) is then applied to formulate an algorithm that performs fast, adaptive and state-dependent RSA, CAC and GoS control. The proposed algorithm can easily be integrated with existing protocols for fast connection setup. Moreover, we discuss the cost efficiency of selected network implementations which are based on basic, colorless and colorless and directionless optical nodes.