Flow-optimized random access for wireless multihop networks

We study elastic traffic performance in time-slotted wireless random access networks, where the link access probabilities are adjusted at flow level each time a new flow (file transfer) arrives or a flow in progress terminates. By defining the capacity sets of such networks we carry out performance analysis in a dynamic setting, where the flows come and go and share the resources according to the balanced fairness principle. We derive flow throughput analytically for three special scenarios and devise an algorithm for general networks. At low loads, flow-optimized random access yields almost the same performance as optimal/deterministic scheduling.