Over the past few years, tremendous advances have been made in mobile computing and wireless communication technologies, including wireless high-speed networks, portable wireless devices, mobile application standards, and supporting software technologies. As a result, mobile data has been flooding the commercial market recently. However, various constraints of mobile computing environments, such as scarce wireless bandwidth and limited client resources, remain as barriers that need to be overcome before the vision of mobile computing can be fully realized. Thus, sophisticated data management and resource management techniques are needed for the enhancement of the performance of mobile data access. This thesis attempts to address some of the performance issues by applying advanced client-side data caching techniques.; We first consider general wireless data dissemination services and propose a cache replacement policy, called Min-SAUD, concerning a realistic mobile environment. The Min-SAUD policy takes into account the cost of ensuring cache consistency before each cached item is used. In addition, Min-SAUD considers access probability, update frequency, retrieval delay, and data size in developing the cost function which determines the cached item(s) to be replaced. The analytical study we performed shows that Min-SAUD achieves the best access performance in terms of stretch, which is a widely used metric for variable-size data.; Next, we concentrate on the caching issues for an emerging mobile data application, i.e., location-dependent information services (LDISs). In LDISs, the server returns different results to a query when it is issued by clients at different locations. Since the location of a mobile client changes constantly, location-dependent data cached at the client may become obsolete, not only due to updates on data items but also because of client movements. Previous work on cache invalidation issues focused on data updates only. In this thesis, we consider data inconsistency caused by client movements and propose several location-dependent cache invalidation schemes for both symbolic and geometric location models.; We also envisage that the spatial property of LDISs brings new challenges for cache replacement policies. In LDISs, the chance of a cached data instance being used again depends on the size of its valid scope, which is defined as the area within which the data instance remains valid. Based on this observation, we propose two novel cache replacement policies for location-dependent data, called PA and PAID. Unlike conventional replacement policies, the PA and PAID policies take into consideration the valid scope area of a data instance. Moreover, they account for the factors of access probability (for both) and data distance (for PAID only). A series of simulation experiments are conducted to evaluate the proposed caching strategies.
展开▼
