Search

Efficient application scheduling with low energy consumption is critical for achieving high performance in heterogeneous computing system (HCS). One of the classical approaches is the combination of heuristics and Dynamic Voltage Scaling (DVS). We propose a new schedule scheme which concentrates on the energy consumption of supply voltage's varying and a modified algorithm which analyses the extra energy cost by voltage switches and the energy saving by working at a lower level to get a better energy performance of the whole system. The performance comparison study shows that the proposed algorithm outperforms the existing scheduling algorithms. [ABSTRACT FROM AUTHOR]

Mobile to mobile video conferencing is one of the services that the newest mobile network operators can offer to users. With the apparition of the distributed video coding paradigm which moves the majority of complexity from the encoder to the decoder, this offering can be achieved by introducing a transcoder. This device has to convert from the distributed video coding paradigm to traditional video coding such as H.264/AVC which is formed by simpler decoders and more complex encoders, and allows to the users to execute only the low complex algorithms. In order to deal with this high complex video transcoder, this paper introduces a graphics processing unit based transcoder as base station. The use of graphic accelerators in this framework has not been proposed before in the literature and offer a new field to explore with promising results. The proposed transcoder offers a time reduction of the whole process over 79% with negligible rate distortion penalty. [ABSTRACT FROM AUTHOR]

This article presents information on heterogeneous distributed computing. A metatask is a collection of independent tasks that need to be mapped onto a heterogeneous computing (HC) suite. Some tasks may have subtasks with data dependencies among them. Most of the heuristics and environments considered in the article are suitable for mapping tasks that can be decomposed into subtasks with data dependencies. The on-line approach assigns each task to a machine when it is submitted. The interval approach waits for a set of new tasks to arrive and then maps those tasks and remaps any earlier tasks that have not yet started execution. Developing heuristics for matching and scheduling metatasks in HC systems is an active research area.

A definition of the term "heterogeneous system" is presented. In software context, heterogeneous systems refer to systems that have a different aspects such as the interface, the implementation and the data. Two systems in a family are heterogeneous to the extent that they are incompatible in some way.