A grid workflow-based Monte Carlo simulation environment
Yaohang Li, Michael Mascagni, Robert van Engelen, Qin Cai · 2004
Monte Carlo methods provide enormous scope for realistic statistical modeling and simulation. The implementation of large-scale Monte Carlo applications on the grid benefits from state-of-the-art approaches to accessing resources in a computational grid. Workflow techniques allow one to describe and enact his simulation processes in a structured, manageable, and verifiable way. We developed the Grid-Computing Infrastructure for Monte Carlo Applications (GCIMCA) based on the Globus toolkit and the SPRNG library. The Globus toolkit facilitates the creation and utilization of a computational grid for large distributed computational jobs and the Scalable Parallel Random Number Generators (SPRNG) library is designed to generate practically infinite number of random number streams with favorable statistical properties for parallel and distributed Monte Carlo applications. GCIMCA provides grid services specific to gridbased Monte Carlo simulation applications, including the Monte Carlo subtask schedule service using the N-out-of-M strategy, the facilities of application-level checkpointing, the partial result validation service, and the intermediate value validation service. By taking advantage of emerging grid workflow paradigms and the facilities of GCIMCA, we implemented a Grid Workflow-based Monte Carlo (GWMC) simulation environment. Workflow management services are implemented to manage the Monte Carlo simulation process. Based on these services, we intend to provide a trustworthy and manageable grid-computing environment for large-scale and high-performance distributed Monte Carlo simulation applications. 1 1.