CAREER: Adaptive Resource Management in Highly Dynamic Real-Time Systems with Physical Constraints

University Of Illinois At Urbana-Champaign, Urbana IL

Investigators

Abstract

Caccamo 0237884 This CAREER project is a combined research and educational project in the area of resource management for highly dynamic real-time systems with physical constraints. The new millennium heralds the convergence of communication, computing, and intelligent control of the physical environment. The rapid advancement of computing and wireless technologies will enable us to employ cooperative real-time nodes that must operate without human intervention. These nodes frequently reside in hostile environments for accomplishing tasks ranging from space monitoring and surveillance to environmental protection. Under this challenging vision, there will be an extensive deployment of highly dynamic and physically constrained real-time node, connected together through communication networks. This project is aimed at the development of new resource management techniques for dynamic, constrained real-time systems. The research follows four avenues: 1) Collaborative scheduling: Distributed rate adaptation and collaborative resource reclaiming techniques are explored to mitigate the effects of higly dynamic workloads in a distributed system with tightly-coupled real-time collaborative tasks running on different nodes. 2) Prioritized medium access control (MAC) with rate-adaptive messages: Traditional MAC protocols are not suitable for real-time nodes collaborating through wireless networks because messages exchanged inside the network are largely periodic and need guaranteed bounded delay. This research considers how to prioritize access to guarantee bounded delay, and how to implement distributed rate adaptation. 3) Power management: The research explores how to exploit the predictable schedule of the wireless channel as an effective way to limit power consumption using sleep mode. 4) Template design: Off-line template design will be explored for managing multiple physical constraints such as energy and thermal bounds in real-time CPU and network resource scheduling. The broader impacts of this CAREER project are both technical and educational. The theory and techniques developed in this project will have a broad impact, as they enable a wide spectrum of critical real-time applications. The educational goals of the project are to transfer the research results to undergraduate and graduate courses, industry-sponsored workshops, and the national security and commercial sectors. The educational plan is aimed at providing undergraduate and graduate students knowledge and skills to ensure their technical leadership in the real-time community. This plan is realized by revising an existing undergraduate course and introducing a new graduate course, as well as developing widely-available Open Source toolkits and curriculum materials for research and educational use.