The public defence of Filip Markovic’s licentiate thesis in Computer Science and Engineering

Datum: 2018-09-21
Tid: 13.15
Plats: room Kappa, MDH Västerås.

The public defence of Filip Markovic's licentiate thesis in Computer Science and Engineering will take place at Mälardalen University on September 21, 2018, at 13.15 PM in room Kappa, Västerås.

Title: “Improving the Schedulability of Real Time Systems under Fixed Preemption Point Scheduling”.  

Serial number: 270.

The examining committee consists of Assistant Professor Sebastian Altmeyer, University of Amsterdam; Associate Professor José Javier Gutiérrez, University of Cantabria; Associate Professor Christian Berger, Chalmers University of Technology; Among the members of the examining committee, Assistant Professor Sebastian Altmeyer has been appointed the faculty examiner.

Reserve; Associate Professor Moris Behnam, Mälardalen University.

 

Abstract:

In some computer systems, it is not only important to produce the correct result in each activity (e.g. computing that 2+2=4), but it is also important that activities are finished within a specified time, e.g. notifying the personnel in a nuclear power plant about a radiation leakage. If the leakage is not detected and the notification sent within a minute from the moment it happened, then the life of the personnel can be severely endangered. In such systems it is crucial to guarantee timeliness, meaning that we can guarantee that all of the activities that need to be performed within a specified time, will really be, regardless of the situation. For this purpose, we use scheduling, which is the mechanism that controls in what order different activities are carried out. Also, in order to test whether a certain scheduling can guarantee that no timing constraints are violated we use schedulability analysis.

In this work, we present three new ways to improve schedulability analysis for a particular type of systems by: i) dividing activities in a clever way when more than one processor is available; ii) making detailed analysis of the time wasted when activities interrupt each other; and iii) utilizing probabilistic information about the expected interruption penalties. Together, the proposed approaches result in more precise analysis and better use of scarce computational resources.