Analysis of a Mobile - Assisted Adaptive

Source messages intended for a mobile host can be routed in one of two ways. Either the source knows the direct route to the mobile host, and is informed of all location changes by the mobile host (informed routing), or the source directs messages to a home agent that forwards messages to the mobile host (triangle routing). When the rate at which the mobile host changes location and the rate at which messages are directed to the mobile host are known and xed, we show that the optimal routing policy is described by a threshold rule that depends on the normalized diierential route cost between triangle and informed routing and the call to mobility ratio (source messaging rate divided by location update rate). Since the call to mobility ratio may not be known a priori or may change slowly with time, we also derive an adaptive policy selection algorithm. The policy is derived from a maximum likelihood estimate of the call to mobility ratio based on observations of message arrivals and location changes. The algorithm is found to work well when there is a clear advantage to either triangle or informed routing. However, when the two routing schemes are relatively close in average cost, the algorithm performance is degraded by repeated policy reversals. For this reason, algorithms which use hysteresis and/or a preset preference (preference threshold) for one routing scheme or another were explored. It was found that neither hysteresis, nor preference threshold techniques alone performed well, but rather a combination of the two resulted in greatly improved performance for a wide range of values of the call to mobility ratio.