On the cost of global knowledge in ad-hoc networks

Locally available information such as local load, battery status or neighbor degree might not lead to optimal or even near-optimal results when used for decision processes such as routing. Global knowledge would enable a node to evaluate its relative state by comparing its local value against the global value of the respective metric. To be able to evaluate the relative state correctly a network-wide, global view has to be calculated at each node in the network. Each node can base local decisions on that knowledge to achieve global objectives such as load balancing in a cross-layer fashion. This way, expensive network-wide operations can be avoided. This paper analyzes the cost involved to locally establish such global knowledge in terms of memory and communication overhead without generating periodic control messages or flooding the network. It evaluates different caching strategies and a novel data dissemination technique that is part of the crosstalk cross-layer framework.