Abstract
The growing computational power requirements of grand
challenge applications has promoted the need for linking high-performance
computational resources distributed across
multiple organisations. This is fueled by the availability of the
Internet as a ubiquitous commodity communication media,
low cost high-performance machines such as clusters across
multiple organisations, and the rise of scientific problems of
multi-organisational interest. The availability of expensive,
special class of scientific instruments or devices and data
sources in few organisations has increased the interest in
offering a remote access to these resources. The recent
popularity of coupling (local and remote) computational
resources, special class of scientific instruments, and data
sources across the Internet for solving problems has led to the
emergence of a new platform called Computational Grid.
This paper identifies the issues in resource management
and scheduling driven by computational economy in the
emerging grid computing context. They also apply to clusters
of clusters environment (known as federated clusters or
hyperclusters) formed by coupling multiple (geographically
distributed) clusters located in the same or different
organisations. We discuss our current work on the Nimrod/G
resource broker, whose scheduling mechanism is driven by a
user supplied application deadline and a resource access
budget. However, current Grid access frameworks do not
provide the dynamic resource trading services that are required
to facilitate flexible application scheduling. In order to
overcome this limitation, we have proposed an infrastructure
called GRid Architecture for Computational Economy
(GRACE). In this paper we present the motivations for grid
computing, resource management architecture, Nimrod/G
resource broker, computational economy, and GRACE
infrastructure and its APIs along with future work.