Experimental Evaluation of Drift: Total Message Ordering in Ad Hoc Networks

We present DRIFT1, a total ordering multicast algorithm optimized for ad hoc networks. DRIFT combines virtual flooding with a communication history ordering algorithm. Virtual flooding is a way of using unrelated message streams to propagate message causality information in order to accelerate message delivery. We review the total order multicast problem and the unique challenges posed in ad hoc networks. We describe DRIFT and its use of virtual flooding in detail. We provide optimizations for implementation. We implement DRIFT on a wireless sensor network. For comparison, we implement the communication history ordering algorithm, TOF, on which DRIFT is based. Both require that all messages be received by all nodes, we implement a reliable local broadcast scheme to achieve reliable message transmission. We evaluate DRIFT’s performance through experimentation and simulation. We employ wireless channel emulation as well as radio transmission in our experimental approach. We study the effects of varying the relative rate at which messages are multicast using both physically flooded messages and nonflooded messages as carriers of the virtual flood. We examine the impact of varying the period of non-flooded messages. From our wireless transmission experiments we find DRIFT delivers multicast messages twice as fast as TOF using only physically flooded messages as carriers. DRIFT achieves a sixfold speedup when periodic non-flooded messages act as carriers in addition to the physically flooded messages.