| Resumo: | Computing services is a growing industry in the last decade with a increasingly broader audience. Scientific organizations use these giant computational infrastructures to provide their collaborators computational power that their personal computers don't have, reducing task execution times like satellite photos compilation from several days to a few hours. In order to house more services and have shorter responses times it's necessary to study system flaws and primary delayers and fix them. This report targets task scheduling in HPC networks and studies virtualization technologies and the problem of interference between virtual machines. Through task allocation in hosts that minimize interference effects, execution time reduction allow the same resources to execute a bigger number of tasks in the same period of time. For demonstration purposes, the SimGrid platform is used to compare results of the proposed algorithm against standard industry algorithms.
|
|---|