Resource management for large real-time systems

点击翻译按钮获取中文摘要

This research program focuses on the theoretical foundations and engineering principles for the design of high-confidence real-time systems.Real-time systems execute many tasks concurrently and need to respond to events within expected deadlines. This requirement calls for careful resource allocation and task sequencing. The increased use of embedded computing and greater interaction with the physical world has led to more systems becoming ``real-time.'' Moreover, even services offered over networks are expected to deliver results on time because of service-level agreements between service providers and clients. There is, however, a lot of dynamic behavior -- because of complexities in system architecture and because of varying workload -- that limits the use of traditional deterministic analysis. This research program will develop a mathematical basis for reasoning about predictable behavior (mainly temporal behavior) when workload can fluctuate and when allocations along multiple resource dimensions -- not just CPU share -- affect response times. Theoretical analysis will be backed up with the implementation of mechanisms for ensuring that resource allocations can be made with fine granularity and accuracy. These mechanisms will be developed at the operating system level and below to provide system architects and developers with the primitives for building robust real-time systems. The combination of theoretical tools and engineering principles can drive down the cost of system building, which is currently extremely high because of resource over-provisioning, extensive debugging and re-engineering and certification costs. The lowering of costs will lead to greater market penetration for pervasive computing applications that integrate computation, communication and control and provides a significant opportunity for Canada to take a lead role in the design and development of embedded systems and ubiquitous services.