Distributed Active Hierarchy Maintenance in MANETs

Creating a hierarchy to meet certain constraints and optimize objective functions can greatly improve scalability, performance and robustness of large networks. However, as constant optimization would result in excessive reconfiguration and signaling overhead, existing hierarchy maintenance protocols are only invoked after detecting constraint violations (e.g., when a node cannot directly reach other nodes in its part of the hierarchy). This passive approach minimizes overhead, but will, over time, result in declining network optimality, particularly in dynamic MANET networks. This paper investigates a new active approach to maintaining a hierarchy, which maintains better network optimality, while balancing the reconfiguration and signaling overhead. Our active approach is performed only at border routers that, at random intervals, perform a local optimization based on the objective functions globally optimized during the hierarchy generation. Using an example of creating balanced size domains (clusters), our simulation results show that active localized decisions can maintain the hierarchy close to the optimal value even after much network dynamics; this is in marked contrast to passive local maintenance which allows the network to deteriorate dramatically.