Monitoring and Steering of Large-Scale Distributed Simulations
George A. Geist, J.A. Kohl · University of North Texas Digital Library (University of North Texas) · 1999
Oak Ridge National Laboratory is developing a state-of-the-art parallel application development system called CUMULVS, which allows scientists to easily incorporate interactive visualization, computational steering and fault tolerance into distributed software applications. The system is a valuable tool for many large scientific applications because it enables the scientist to visually monitor large data fields and remotely control parameters inside a running application. Collaborative monitoring is provided by allowing multiple researchers to simultaneously attach to a simulation, each controlling their own view of the same or different data fields within the simulation. By supporting steering of a simulation while it is running, CUMULVS provides the opportunity to accelerate the process of scientific discovery. CUMULVS also provides a simple mechanism to incorporate automatic checkpointing and heterogeneous task migration into large applications so that simulations can continue to run for weeks unattended. This paper will give an overview of the CUMULVS system and its capabilities, including several case histories. The status of the project is described with instructions on how to obtain the software.