Output details

A Survey of Hard Real-Time Scheduling for Multiprocessor Systems

<07>The widespread availability of multiprocessor and multi-core platforms has had a profound impact on real-time systems research, in particularly on scheduling theory. Almost none of the foundational results that have been developed over 30 years for single processor systems are applicable. This has led to an intense period of research in the last decade. Although extensive, much of this research has been unstructured, isolated and lacking in common terminology, notations, models and methods. This survey paper provides a classification and taxonomy of techniques used for multiprocessor schedulability analysis. It contains a rigorous review and critique of over 150 papers that appeared in the research literature prior to the end of 2009. It includes sections on metrics that can be used to compare different scheduling algorithms and fundamental results that are independent of the algorithm used. It has facilitated a more focused international research effort in this critically important area. The importance of this is both theoretical and practical (as most deployed embedded systems are now multi-core). This survey has attracted considerable attention in the community. The article sets out the remaining challenges and key open issues for the research community to address. For example, it identifies semi-partitioned approaches as potentially providing an effective bridge/compromise between fully partitioned and fully global approaches to scheduling. Subsequent research has substantiated this observation. Although aimed at the research community, for new PhD students the survey has proved successful in providing a thorough introduction and overview of the field. It is listed as a seminal work recommended for education by the IEEE Technical Committee on Real-Time (http://tcrts.org/education/seminal-papers/).