ERTOS Public Seminar
Real-time Scheduling on Multiprocessors
Dr. Björn Andersson, Institute Polytechnic Porto, Portugal
Time/Venue
23rd August 2006, 5:30pm
National ICT Australia Ltd, Level 1 Seminar Room, 223 Anzac Parade (Building L5), Kensington NSW 2052
Abstract
Multiprocessors have already made the transition from high-end computing to desktops and laptops. This was possible because of the miniaturization of integrated electronics system which allowed the implementation of multiprocessors on a single chip, called multicores. Now, the next step is about to begin. These multicores are targeting embedded real-time systems as witnessed by (i) the commercial availability of multicore PowerPC processors and (ii) Intel´s recent marketing of the use of multicores in embedded systems.
Unfortunately, the real-time technologies as we know of so far are not well prepared. Real-time scheduling algorithms for single processors perform miserably when applied to multiprocessors as the number of processors increase. The current state of art leaves designers with two choices (i) algorithms that require very frequent context switching or (ii) algorithms that can miss deadline although only 50% of the processing capacity is requested. In this talk, I will tell the history of the development of multiprocessor scheduling algorithms for real-time systems and explain what makes multiprocessor scheduling challenging still today.
Biography
Björn Andersson celebrated his 13th birthday on Friday the 13th in the third month of the year and he has been lucky after that! He received his M.Sc. degree in electrical engineering at Chalmers University of Technology in Sweden and his Ph.D. degree in computer engineering at Chalmers as well. He visited University of Virginia during 6 months and he currently holds the highest post of visiting scientist ("Cientista Convidado") at Institute Polytechnic Porto.
His research interest focus on designing algorithms for sharing resources in real-time systems and proving their performance. The research is theoretical but the results are tested in order to test their validity when he has doubts in their validity. In his Ph.D. thesis he studied multiprocessors. Now he is studying medium access control protocols for wireless networks with real-time requirements, with focus on sensor networks.