Author:
Michael Short
Affiliation:
University of Leicester, United Kingdom
Keyword(s):
Embedded Systems, Non-preemptive scheduling, Feasibility Analysis, Complexity.
Related
Ontology
Subjects/Areas/Topics:
Artificial Intelligence
;
Computer and Microprocessor-Based Control
;
Formal Methods
;
Informatics in Control, Automation and Robotics
;
Intelligent Control Systems and Optimization
;
Planning and Scheduling
;
Real-Time Systems Control
;
Signal Processing, Sensors, Systems Modeling and Control
;
Simulation and Modeling
;
Symbolic Systems
Abstract:
Non-preemptive schedulers, despite their many perceived drawbacks, remain a very popular choice for practitioners of real-time and embedded systems. Although feasibility conditions for non-preemptive scheduling models have previously been considered in the literature, to date little attention has been paid to the non-preemptive ‘thrift’ (or ‘TTC’) cyclic scheduler. This type of scheduler differs from a standard ‘cyclic executive’ in that it does not allow the use of inserted idle-time, and it does not require a lookup table of task executions over the major cycle of the schedule; a feasible schedule is effectively created by assigning release times (‘offsets’) to the tasks. To this end, this paper seeks to address the complexity of generating a feasible cyclic schedule for such a scheduler. It will be shown that when a single set of release times is assigned to the tasks, deciding feasibility of the resulting schedule is coNP-Complete (in the strong sense); and the release time assig
nment problem for such a scheduler is complete for ∑2p.
(More)