Skip to Main content Skip to Navigation
Theses

Ordonnancement de tâches efficace et à complexité maîtrisée pour des systèmes temps-réel

Abstract : The performance of scheduling algorithm influences the performance of the whole system. Real time scheduling algorithms have theoretical optimal schedulable bounds, but these optimal bounds are achieved at the cost of increased scheduling events (preemptions and migrations of tasks) and high run time complexity of algorithms. We believe that by exploiting parameters of tasks, these algorithms can be made more efficient and cost conscious to increase Quality of Service (QoS) of application. In first part of thesis, we propose mono-processor scheduling algorithms which increase quality of service of hybrid application by maximizing execution of soft real time tasks, and by providing guarantees to hard real time tasks even in overload situations. Scheduling cost of algorithms is also reduced (in terms of reduced number of preemptions) by taking into account all explicit and implicit parameters of tasks. Reducing scheduling overheads not only increases performance of the scheduler but also minimizes energy consumption of the system. That's why, we propose to devise a technique embedded with existing DVFS (dynamic voltage and frequency scaling) techniques to minimize the switching points, as switching from one frequency to another steals processor cycles and consumes energy of system. Multiprocessor scheduling algorithms based on fluid scheduling model (notion of fairness), achieve optimal schedulable bounds; but fairness is guaranteed at the cost of unrealistic assumptions, and by increasing preemptions and migrations of tasks to a great extent. An algorithm (ASEDZL) is proposed in this dissertation, which is not based on fluid scheduling model. It not only minimizes preemptions and migrations of tasks but relaxes the assumptions also due to not being based on fairness notion. Moreover, ASEDZL is also proposed to schedule tasks in hierarchical approach, and it gives better results than other approaches.
Document type :
Theses
Complete list of metadatas

Cited literature [74 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00454616
Contributor : Sophie Gaffé-Clément <>
Submitted on : Tuesday, February 9, 2010 - 7:59:54 AM
Last modification on : Monday, October 12, 2020 - 10:27:28 AM
Long-term archiving on: : Friday, June 18, 2010 - 7:38:36 PM

Identifiers

  • HAL Id : tel-00454616, version 1

Collections

Citation

F. Muhammad. Ordonnancement de tâches efficace et à complexité maîtrisée pour des systèmes temps-réel. Autre. Université Nice Sophia Antipolis, 2009. Français. ⟨tel-00454616⟩

Share

Metrics

Record views

499

Files downloads

2359