Panel on Real-Time Scheduling: Where are we now and where are we going? Sang H. Son Computer Science Department University of Virginia Charlottesville, VA 22903, USA
[email protected]
Fax: (804) 982-2214 Chair:
Sang H. Son
Panelists:
Kwei-Jay Lin (University of California at Imine, USA) Jane Liu (University of Illinois, USA) Hide Tokuda (Carnegie Mellon Univ., USA) Pat Watson (IBM Federal Systems Company, USA) Horst Wedde (Wayne State University, USA) Wei Zhao (Texas A&M University, USA)
In real-time systems, tasks have explicit timing constraints such as deadlines. The correctness of the system depends not only on the logical results but also on the time at which the results are generated. Satisfying the timing requirements demands the scheduling of system resources according to wellunderstood algorithms so that the timing behavior of the system is understandable and predictable. Traditionally, many real-time systems use cyclic executives to schedule multiple threads of execution. Under this approach, an application programmer determines the execution timeline by hand to senalize the execution of critical sections and to meet the deadlines. While such an approach is manageable for simple systems, it is not scalable to large systems. The objective of the real-time scheduling theory is to provide a foundation for developing scheduling algorithms that can determine, for a given set of real-time tasks, whether a schedule for executing the tasks exists such that the timing, precedence, and resource constraints of the tasks are satisfied. A scheduling algorithm is said to be optimal if, for any set of tasks, it always finds a schedule that satisfies the constraints of the tasks, if such a schedule exists. Since the introduction of the rate-monotonic scheduling algorithm, a significant amount of work has been performed in the area of real-time scheduling. Several scheduling problems have been solved, a few optimal scheduling algorithms found, and schedulability conditions developed. Can we make a claim that the area of real-time scheduling is now mature? Unfortunately, the answer seems not clearly positive. It would be fair to state that many real-time scheduling issues are well understood, but there are lots of new challenges ahead of us. For example, we have to deal with large distributed real-time systems where end-to-end scheduling guarantee is necessary. Some of the nodes in such a system may utilize parallel architectures, Some of the tasks are not periodic and their resource requirements may not be static. A well-known single-CPU scheduling algorithm for periodic tasks with static priority cannot solve such a problem. In this panel, we would like to address some of the real-time scheduling issues, and discuss about new challenges. The issues will include: What are the important contributions? What are the remaining problems and how to address them? Are the solutions scalable? Are the models appropriate?
132 1068-3070/93 $3.W Q 1993 IEEE
