Date Friday, April 22, 2011, 11:00 AM - 12:00 PM, SB 113
Title: Allocation, Isolation, & Predictability for Compositional Real-Time Systems
Speaker: Nathan Fisher
Many safety-critical systems such as automotive or avionic systems have strict timing requirements. Due to their inherent complexity and size, component-based design techniques are required for decomposing such systems into smaller, simpler components. Until recently, most component-based design techniques have focused primarily on satisfying the functional requirements of the system and have largely often ignored non-functional requirements such as real-time timing constraints. To address the need for validation of timing constraints within component-based design, recently-proposed compositional frameworks permit the abstraction of temporal requirements of a component (and potential sub-components) via relatively simple real-time interfaces. However, three fundamental challenges arise from such abstractions: 1) Allocation: how do we determine the optimal interface which accurately describes the temporal requirements of a component? 2) Isolation: how can components be isolated from the temporal faults or misbehavior of other components? 3) Predictability: in the event of a temporal fault in one component, can we accurately predict which other components will be affected?
In this talk, I describe our recent work on the above challenges via the development of efficient schedulability analysis that provides formal guarantees on the correctness of compositional real-time systems. Furthermore, each of our solutions is either provably optimal or has bounded deviation from the optimal solution. Finally, I will discuss some surprising implications of our obtained solutions for compositional systems that arise in the design of thermal-aware real-time systems.
Nathan Fisher is an Assistant Professor in the Department of Computer Science at Wayne State University. He received his Ph.D. from the University of North Carolina at Chapel Hill in 2007, his M.S. degree from Columbia University in 2002, and his B.S. degree from the University of Minnesota in 1999, all in computer science. His research interests are in real-time and embedded computer systems, parallel and distributed algorithms, resource allocation, and approximation algorithms. His current research focus is on multiprocessor scheduling theory, composability of real-time applications, and thermal-aware real-time system design. He received the NSF CAREER Award in 2010 and multiple best paper awards for his research on the topics of multiprocessor real-time scheduling and compositional real-time systems