Resource-eecient Policies for Information Transfer in a Mobile Environment

Mobile computing environments are traditionally resource-constrained with mobile computers typically characterized by less network bandwidth and power availability than stationary computers. Ideally, users should be able to experience both mobility and performance without sacriicing one for another. Therefore, there is a need for information retrieval policies that are resource eecient. In this work we propose policies for doing resource-eecient information transfer. The basis of the proposed policies is that information users tend to have diierent levels of interest for components within an information element. The new policies we propose use a user proole in ltering out the \use-less" or \uninteresting" information components, to minimize the amount of information transferred and processed. The work focuses on an implementation for the transfer of network news to minimize wireless bandwidth and mobile computer power use. The implementation was done through modiications to the xrn newsreader and Network News Transfer Protocol (NNTP) for a lter process to transfer individual article components (such as headers or screens of the article body) to the mobile host. The transfer policies are evaluated individually and in combination with respect to network bandwidth and the number of requests needed for the newsreader to obtain components of a article. Behavior of the policies with respect to information transfer latency is also evaluated. The results from testing show that for a variety of users and article sizes, user proole-based component-level ltering policies yield better overall performance than conventional transfer methods.