|
Abstract : |
The explosive growth in wireless communications systems and the demand for advanced mobility features have created novel data management problems. Current schemes to address these problems rely on database organizations that have limited functionality and performance anomalies. We propose a new data management scheme that is flexible and scalable, and that is incrementally deployable so it can coexist with current data management standards. We compare our protocol against current standards and other suggested protocols using realistic calling and mobility patterns. To do so, we have built Pleiades, an extensible event driven simulator that is easily configurable to arbitrary geographies, networks, calling and mobility patterns, and data management schemes. We present, for the first time, models to closely approximate user calling and mobility patterns. These models are validated against real call traffic and urban vehicle data. Based on simulations for a representative 24-hour period in the San Francisco Bay Area, we report on several performance measures: global and local database transaction load in terms of number of reads and writes, and network activity in terms of number of messages and average number of message hops. 1, |
