A Framework for Interactive Parameter Sweep Applications

Abstract

Summary form only given. A typical parameter sweep application (PSA) is a parameterized application which has to be executed independently large number of times, to locate a particular point in the parameter space that satisfies certain criteria. From the perspective of domain scientists, the complexity of underlying grid environment should be hidden so that domain scientists can focus on their main concern on performing their experiments. The system needs to provide friendly user environment for scientists to change parameter space or set new policy for execution at runtime. The framework should provide flexible interface for porting legacy applications to a PSA. We proposed five functional components for the interactive framework: a GUI for user to describe experiment, a visualizer to presenting computing results, a coordinator to schedule the execution of computing tasks, a repository to collect computing results, and interface to job scheduling tools, e.g., from Grid. In the current prototype, a tuple space like workspace is used to maintain state of experiment. This prototype allows basic interactivity such as modification of experiment state during execution which demonstrated the capability to manipulate parameter sweep tasks at run time.