ITR: Supporting Adaptable Pervasive Computing Through a Kernel-Middleware eXchange

点击翻译按钮获取中文摘要

McKinleyThe rapid expansion of mobile computing infrastructure promises ubiquitous access to the Internet. However, the dynamics of this environment (variable conditions on wireless networks, changing security policies as users roam among wireless domains, and computing priorities that depend on battery life) pose a significant challenge in the design of software infrastructure for mobile systems. One approach to this problem is to introduce a layer of adaptive middleware between applications and underlying transport services. While a traditional role of middleware is to insulate application components from platform variations, an adaptive middleware layer should exploit services provided by the platform operating system to enhance functionality and performance. This project investigates the interaction between adaptive middleware and the operating system kernel. Specifically, the project focuses on the question: What services and interfaces should the operating system provide to enable adaptive middleware to better meet the needs of mobile computing applications? The key hypothesis to be tested is that this relationship is bidirectional: certain middleware functionality is more effective when pushed into the kernel, and certain kernel operations are more effective when combined with middleware-level information. The Kernel-Middleware eXchange (KMX) project builds on the idea that, instead of viewing lower-level middleware as an "operating system abstraction layer," it is also necessary to investigate how this part of the middleware might actively cooperate with the operating system in order to provide applications with supporting functionality that neither part could provide by itself. The project investigates three cross-cutting concerns: quality-of-service of multimedia communication on wireless networks; measures to enhance the security of mobile devices, including adaptive auditing and packet filtering; and managing battery lifetime through data transformations, adaptive process scheduling, and management of network interface card usage.