Demand Allocation Without Wavelength Conversion Under a Sliding Scheduled Traffic Model

It has been shown that for the scheduled traffic model, connection holding time aware algorithms lead to more efficient resource allocation. The setup and teardown times of the scheduled demands may be fixed, or may be allowed to slide within a larger window. A number of optimal integer linear program (ILP) solutions for the first problem (fixed setup/teardown times) have been presented in the literature, for wavelength convertible networks. In this paper we present a new and complete ILP formulation for both fixed window model, and the more general sliding scheduled traffic model, where the setup and teardown times may vary within a specified range. We consider fault-free as well as survivable networks using path protection, and do not require any wavelength conversion. Our ILP can jointly optimize the problem of scheduling the demands (in time) and allocating resources for the scheduled lightpaths. We have shown that the complexity of our formulation for sliding scheduled traffic model, in terms of the number of integer variables, is less than existing ILP formulations for the simpler fixed window model. For very large networks, we have proposed a fast two-step optimization process. The first step schedules the demands optimally in time, such that the amount of overlap is minimized. The second step uses a connection holding time aware heuristic to perform routing and wavelength assignment for the scheduled demands.