A framework for evaluating network optimization techniques

There is little understanding about how network applications can benefit from mathematical/heuristic optimization techniques, such as cross-layer design. We propose CLASS, a component-based framework for evaluating and comparing network optimization techniques, such as cross-layer design. CLASS uniquely integrates network optimization with high-fidelity simulation to provide an “optimize-and-verify” loop to network designers. Within CLASS framework, we propose two optimization algorithms. One algorithm (ROCA) jointly optimizes routing and channel assignment for cognitive-radio networks. The other algorithm (Flow-Aware Routing or FAR) jointly optimizes routes and MAC schedules in single-channel networks. Both algorithms exploit spatial diversity by reusing channels or slots. We discuss the benefits and applicability of both algorithms in the context of surveillance applications.