Providing Real-Time Applications With Graceful Degradation of QoS and Fault Tolerance According to$(m, k)$-Firm Model

J. Li, Y. Song, F. Simonot-Lion
2006 IEEE Transactions on Industrial Informatics  
The ( )-firm model has recently drawn a lot of attention. It provides a flexible real-time system with graceful degradation of the quality of service (QoS), thus achieving the fault tolerance in case of system overload. In this paper, we focus on the distance-based priority (DBP) algorithm as it presents the interesting feature of dynamically assigning the priorities according to the system's current state (QoS-aware scheduling). However, DBP cannot readily be used for systems requiring a
more » ... inistic ( )-firm guarantee since the schedulability analysis was not done in the original proposition. In this paper, a sufficient schedulability condition is given to deterministically guarantee a set of periodic or sporadic activities (jobs) sharing a common non-preemptive server. This condition is applied to two case studies showing its practical usefulness for both bandwidth dimensioning of the communication system providing graceful degradation of QoS and the task scheduling in an in-vehicle embedded system allowing fault tolerance. Index Terms-( )-firm, non-preemptive scheduling, quality of service (QoS), real-time. . His research interests include, on the one hand, the modeling and performance evaluation of networks and real-time distributed systems using queueing analysis, network calculus, and scheduling theory, and on the other hand, the implementation of real-time QoS mechanisms in fieldbuses, in-vehicle networks, switched Ethernet, IP networks, and power line communication networks. Françoise Simonot-Lion is a Professor of computer science at the Institute National Polytechnique de Lorraine, Nancy, France. Her main research topics are modeling and verification techniques for the design of optimized real-time distributed applications under safety constraints as well as specification of embedded services ensuring a real-time quality of service (scheduling of tasks and messages, real-time middleware, and frame packing).
doi:10.1109/tii.2006.875511 fatcat:5x3maqu7rvh7bnt64fdcxfxsym