Abstract

Computational Grids are becoming attractive and promising platforms for solving large-scale (problem solving) applications of multi-institutional interest. However, the management of resources and scheduling computations in the Grid environment is a complex undertaking as they are (geographically) distributed, heterogeneous in nature, owned by different individuals or organisations with their own policies, different access and cost models, and have dynamically varying loads and availability. This introduces a number of challenging issues such as site autonomy, heterogeneous substrate, policy extensibility, resource allocation or co-allocation, online control, scalability, transparency, and economy of computations. Some of these issues are being addressed by system-level Grid middleware toolkits such as Globus.

Our work in general focuses on economy/market driven resource management architecture for the Grid; and in particular on resource brokering and scheduling through a user-level middleware system called Nimrod/G and economy of computations through a system-level middleware infrastructure called GRACE (GRid Architecture for Computational Economy). Nimrod/G supports modeling of a large-scale parameter study simulations (parameter sweep applications) through a simple declarative language or GUI and their seamless execution on global computational Grids. It uses GRACE services for identifying and negotiating low cost access to computational resources. The Nimrod/G adaptive scheduling algorithms help in minimising the time and/or the cost of computations for user defined constraints. These algorithms are evaluated in different scenarios for their effectiveness for scheduling parameter sweep applications in Grid environments such as GRACE and core middleware (Globus, Legion, and/or Condor-G) enabled federated Grids.