Multicriterial remapping in heterogeneous environment based on multiple server tasks

Heterogeneous environment of renewable, but also timing variable resources like the computational grid, requests special attention by mapping. Regardless of the static and the dynamic phase of mapping, unexpected events in the environment can significantly influence application execution. The predictive-reactive remapping procedure enables significant improvements. It is run after the missed remapping threshold by using a suitable period, horizon and shift. The aperiodic server mechanism enables successful execution of periodic and aperiodic parts of applications. The remapping procedure based on multiple server tasks enables the usage of one's own resources, but also a fast connection to machines of appropriate performances that will soften the influence of disturbances. Preliminary results indicate possible applications in a broader environment and increasing autonomy.