A Scheduler for Grid Task Based on Differential Evolution Algorithm and Robust to Uncertainty Communication Demand of the Application

The grid schedulers are responsible for deciding how to divide the tasks on the grid hosts in order to achieve the best possible performance. The lack of user´s knowledge about the applications´ demands do the informations received by the schedulers differ from the actual values that should be considered to achieve nearly optimal scheduling. This paper introduces a robust task scheduler to uncertainties in the description of the communication demands of the application. The robustness of scheduler is evaluated using simulations based on real grid applications. The schedules generated by schedulers robust to uncertainties were compared with schedules generated by schedulers that considers the information input are always correct. The robust uncertainties schedulers presented themselves useful before scenarios with a high degree of uncertainty.