Using application information to drive adaptive grid middleware scheduling decisions

We are in the process of building a dynamic, adaptive scheduler for parallel Grid programs, which will form part of the SymGrid-Par middleware for execution of symbolic programs on the Grid. This scheduler will dynamically interact with the applications it controls to obtain foreknowledge of application demands for resources. By monitoring the previous behaviour of the application, it will also be able to obtain foresight of the application's future behaviour. Using both foreknowledge and foresight, this "prescient" scheduler will be able to adapt quickly to changing application demands by tuning the amount of resources allocated to applications "on the fly".