Abstract:
Grid computing, a parallel and distributed computing infrastructure via wide-area sharing of
computational resources, has evolved to be a mainstream technology enabling large-scale virtual
organizations. Since the main objective of grid computing is to support resource sharing within
a networked infrastructure, managing resources properly in grid environment is very important.
In most grid systems where submitted tasks are initially placed into a queue due to unavailablity
of required resources, there is no guarantee as to when these tasks will be executed. This policy
may cause problems for time-critical applications. This policy is also problematic for workflowbased
applications where tasks have inter-dependencies. Using Advance Reservation (AR) in grid
systems allows users to secure or guarantee resources prior to executing their jobs. The resource
reservation is a scheduling process that maps tasks on the distributed resources. One of the major
challenges of resource reservation for a workflow-based application is to minimize the delay of execution
of the overall application. In general, the problem of mapping a set of interdependent tasks
on distributed services belongs to a class of problems known as NP-Complete problems. Thus, in
practice, heuristics are most often used to schedule workflow-based applications in grid environments.
In this thesis some properties like slack time of tasks, critical paths etc. of a workflow-based
application have been exploited to provide a resource reservation scheme that gives better results
and supports advance reservation. We demonstrate our claims by conducting a detailed performance
evaluation and comparing with existing system for grid computing.
