This exploratory proposal addresses some of the algorithmic issues of autonomic distributed systems, and explores its applicability to the self-management of large-scale wireless networks. An autonomic system mimics the human body's autonomic nervous system that regulates homeostatic functions without conscious intelligent control, and provides facilities for self-management in large-scale complex heterogeneous systems. The proposal argues that with the rapid growth of wireless networking and cognitive radios, combined with the demand for seamless ubiquitous services, autonomic computing faces interesting challenges and opportunities. It examines various algorithms for self-management, and illustrates their applicability in the large wireless networks providing ubiquitous services. These include methods of stabilizing streaming applications in a wireless network, stabilization of distributed applications containing selfish agents, and designing gracefully degradable systems. It also studies the new paradigm of population protocols that has the potential to generate ambient knowledge in a transparent way in a system of passively mobile agents via fair mutual encounters.
The intellectual merit of the proposal rests on the development of novel foundational tools and techniques for the autonomic management of large distributed systems that are self-stabilized and self-organized and involve node mobility and resource constraints. The proposed research in autonomic distributed systems management is envisioned as potentially transformative as it is likely to boost the growth of next generation wireless networks and the dependability of safety-critical applications that are slated for revolutionary growth in the near future.
The broader Impacts encompass applications of wireless networking leading to increased safety and security leading to more predictable performances, fewer interruptions and better predictability, be it in hospitals, airports, battlefields or highways. The PI will bring the research findings to the classroom and integrate into course curricula, that will become accessible to all graduate students including women and minorities. Results of this research will be disseminated through publications in appropriate conferences and journals. The PI has proven expertise and successful track record in the design of self-managing systems for nearly two decades.
