Middleware support for developing adaptive applications across heterogeneous environments and systems.

详细信息   

• 作者：Wang ; Qiang.
• 学历：Doctor
• 年：2007
• 导师：Cheng, Liang
• 毕业院校：Lehigh University
• 专业：Computer Science.
• ISBN：9780549107767
• CBH：3270673
• Country：USA
• 语种：English
• FileSize：9612519
• Pages：197

文摘

Networked distributed computing environments today are largely heterogeneous, dynamic, and best effort with no guaranteed quality of service, because of the emergence of new applications, diversified computing devices, and the ubiquitous wireless communication. Middleware technology has been used extensively in facilitating the development of distributed applications by abstracting the network and operating system to create a single-system image. In heterogeneous environments, the fundamental design principle of middleware needs to be revisited. It is desirable for distributed applications to be aware of their surroundings and resource-availability so that they can be self-adaptive and be able to adjust their behaviors dynamically to optimize their functions based on current contextual situations.;However, adaptive applications are notoriously difficult to construct since application developers need to have expertise in both the application domain and resource domain. In this dissertation research, AwareWare is developed as middleware that supports adaptive application development. AwareWare employs the middleware-controlled approach, with "adaptation concerns" clearly separated from the functional codes of an adaptive application. AwareWare is fully implemented to provide broad support for both CORBA-based and DLL-based adaptive applications. Case studies and quantitative evaluations have shown that AwareWare provides a complete, flexible, and high-performance solution.;Original research ideas in AwareWare are presented in this dissertation: (1) A novel available bandwidth measurement tool, FEAT, is developed and evaluated, which features a new dynamic pattern of probing packets and provides timely and more accurate results than other current tools; (2) An awareness data distribution/management architecture is proposed, and analytical results show that its hybrid architecture demonstrates significant advantages over the client/server and P2P architectures in terms of flexibility and reduction of wireless resource consumption; (3) A high performance decision engine is designed, which includes an event model, dynamic reconfiguration for three target platforms, and active-message based architecture synchronization; and (4) Reprogramming sensor nodes over-the-air is investigated to facilitate tunable environment awareness, and existing state-of-the-art reprogramming systems are analyzed under a generic reprogramming framework by simulations and model-based analyses.