Resource Allocation Strategies for a non-Continuous Sliding Window Traffic Model in WDM Networks

In recent years, a number of papers have shown that the scheduled traffic model, which exploits knowledge of the connection holding times of traffic demands, can lead to significant improvements in resource utilization in WDM networks. In such a traffic model, the setup and the teardown times of the scheduled demands may be known in advance (fixed window model) or may be allowed to slide within a larger window (sliding window model). In both fixed and sliding window models, once the transmission of a demand is started, it continues until the entire data has been transmitted. However, there are many applications where such continuous data transmission is not strictly required. In this paper, we introduce a new model, the non-continuous sliding window model, where a demand may be decomposed into two or more components and each component can be sent separately. We first present an integer linear program (ILP) formulation for resource allocation under the non-continuous sliding window model and show that both the fixed and the traditional sliding window models can be treated as a special case of our generalized non-continuous sliding window model. Our formulations can accommodate fixed, sliding, and non-continuous demands, or any combination of these demand types. We also provide a heuristic algorithm that can be used for practical sized networks with a large number of overlapping demands. Simulation results on various networks, using different demand sets, show that our model results in significant performance improvements, even over results obtained using traditional scheduled traffic models, which already outperform holding-time-unaware models.