Point of Congestion in Large Buffer Mobile Opportunistic Networks

Abstract

Congestion detection and management are difficult to realize in Mobile Opportunistic Networks (MONs) due to the lack of end-to-end connectivity, global information, absence of ACKs/NACK messages. One of the simplest means of detecting congestion in any network, primarily in the intermediate nodes, is buffer overflow leading to packet drop. Although researchers in this area have used this phenomenon to detect congestion, this is not always true for a MON, especially under large buffers. We show that in MONs with large buffers, there exists a critical buffer occupancy beyond which congestion sets in, and network performance degrades even though the buffer may not be full. To study this behavior, we develop an analytical model based on forwarding probability by considering the nodes to have a large buffer size and finite bandwidth. Using Epidemic forwarding as the underlying routing protocol, the analytical model is used to detect the exact buffer occupancy that leads to congestion. The theoretical results have been compared with the simulation results to prove the correctness of the model.