Responsiveness on the Interactive Grid

The Grid vision promises the secure and dynamic
sharing of heterogeneous resources across the Internet. To date
the emphasis has been on supercomputing scenarios where a
relatively long running job is submitted to the Grid, requiring
little interaction with the end user. More recently, with the
emerging aspirations of Grid aficionados to use it for a much
wider range of purposes – e-commerce and inter-enterprise
computing for example – the issue of using the Grid for
distributed, highly interactive applications has been raised. In
our case this has emerged within a project seeking to build a
European Learning Grid Infrastructure† (ELeGI). Distributed
Learning Environments need to be engaging and responsive, and
as such it is important that, in addition to the dynamic resource
sharing, service-oriented model offered by the Grid, that it can
also meet the responsiveness needs of end users. In this paper
we extend earlier work on analysing delay in distributed
interactive applications to investigate the performance of the
foremost Grid toolkit, Globus, in the context of distributed
learning.