View article

We study the scaling laws for wireless ad hoc networks in which the distribution of n nodes in the network is homogeneous but the traffic they carry is heterogeneous. More specifically, we consider the case in which a given node is the data-gathering sink for k sources sending different information to it, while the rest of the s = n - k nodes participate in unicast sessions with random destinations chosen uniformly. We present a separation theorem for heterogeneous traffic showing that the optimum order throughput capacity can be attained in a wireless network in which traffic classes are distributed uniformly by endowing each node with multiple radios, each operating in a separate orthogonal channel, and by allocating a radio per node to each traffic class. Based on this theorem, we show how this order capacity can be attained for the unicast and data-gathering traffic classes by extending cooperative communication …