Energy Efficient Mapping of Mixed Criticality Applications on Unrelated Heterogeneous Multicore Platforms
Ref: CISTER-TR-160501 Publication Date: 23 to 25, May, 2016
Energy Efficient Mapping of Mixed Criticality Applications on Unrelated Heterogeneous Multicore Platforms
Ref: CISTER-TR-160501 Publication Date: 23 to 25, May, 2016Abstract:
Heterogeneous multicore platforms are becoming
an attractive choice to deploy mixed criticality systems
demanding diverse computational requirements. One of the
major challenges is to efficiently harness the computational
power of these multicore platforms while deploying mixed
criticality applications with timeliness properties. Energy
efficiency is also one of the desired requirements in the
design phase, and therefore it is often difficult for the
system designer to simultaneously satisfy those sometimes
contradictory requirements. In this paper, we propose a
novel partitioning algorithm for unrelated heterogeneous
multicore platforms to map mixed criticality applications.
The algorithm not only ensures the timeliness in different
modes of execution but also tries to allocate the applications
to their energy-wise favourite cores. We considered a realistic
power model that further increases the relevance of the
proposed approach. We have performed an extensive set of
experiments to evaluate the performance of the proposed
approach, and we show that in the best-case, we achieve a
23.8% gain in the average power dissipation over the state-ofthe-art partitioned algorithm. Our proposed algorithm also
has a better weighted schedulability when compared to the
existing partitioned algorithms.
Muhammad Ali Awan
Damien Masson
Eduardo TovarDocument:
View
Download11th IEEE International Symposium on Industrial Embedded Systems (SIES 2016).
Krakow, Poland.
Record Date: 2, May, 2016