Tailoring Java for a Pervasive Service Infrastructure

A growing number of mobile computing devices are becoming available that can access large amounts of data and services over the Internet. While Java appears to be an appropriate platform to deal with diversity, our experience reveals that in a mobile environment it has insufficient support for system facilities such as remote storage, disconnected operation, and concurrent execution of multiple services. We believe that these facilities can be provided transparently and efficiently through the use of API interposition. We investigate the use of interposition to provide access to a remote storage service, to implement a cache for data and services, and to help isolate services that are sharing a Java virtual machine. We describe how our approach is implemented and present some experimental results for CPU and memory usage. Based on our results, we believe that API interposition will make it possible to support these infrastructure features transparently and efficiently with an acceptable overhead