Modeling ad hoc network as queuing network

Abstract

Ad hoc networks are networks in which there is no fixed infrastructure. The mobile nodes are independent and have distributed control. Modeling of Ad hoc network is a critical task, because it includes modeling of mobile behavior of nodes as well as behavior of wireless links. In this work, we propose a novel approach of modeling ad hoc network using mesh connected queuing network. We have shown that a node in Ad hoc network can be modeled as a queue receiving data from DBMAP (Discrete Batch Markovian Arrival Process) sources, which have underlying Markov chain. The packet loss probability of this queue has been simulated and found to match analytical results. The current state of art in modeling mobility in an Ad hoc network consists of selecting a velocity and direction for a node and deriving an analytical model based on these parameters. This approach lacks the concept of path and link availability. In our novel approach, data taken from various mobility scenarios through simulations are used to find out a statistical distribution that gives connectivity between two nodes. The available and unavailable duration distributions obtained from AdHocSim are incorporated into queuing network. The effect of Velocity, Area and Pause time on link availability are studied and presented. Packet error rate and Delay are measured and validated. We show that the proposed model is useful for performance evaluation and parameter tuning of existing routing protocols.