A generic workflow-based model for the deployment problems in grid systems using MPIAB as a case study

Grid Computing is evolving as a novel infrastructure that enables the users to share and efficiently use the compute resources and solve more complex problems. Individuals, organizations, and/or institutions can contribute to form a pool of resources that can be used by everyone based on predetermined policies. Teragrid [18] is one such example that integrates computers, data resources and tools from over nine partner sites spanning across the United States with 102 teraflops of computing capability and 15 petabytes of storage.; The dynamics of Grid environment with such a large pool of resources makes it impossible to determine the specifics of the resources allocated for use such as, the processor type, type of the operating system, and software components available to name a few. These parameters influence the nature and the amount of the computations that can be carried out on a resource. And, the lack of knowledge about these parameters may result in failure of the jobs submitted when a user attempts to perform computations on the resources that do not have the necessary components. This may eventually lead to user manually configuring the resources before use and is not a good solution because user is not assured the same set of resources in a Grid environment with constantly-varying set of resources. We term this problem that requires configuring the Grid resources for a particular type of computation before it can be deployed on them as Deployment Problem and address it in this report.; We propose a service-based solution that identifies the components available on the allocated resources and compares it with the components required to carryout the computation. The comparison yields a set of components missing on the resources allocated but necessary for the computation. Once identified, it dynamically deploys the required components using Karajan workflows. We then present the workflows implemented for deploying software components. We use Message Passing Interface - Agents Based (MPIAB) tasks as an example and present the workflows to detect and deploy MPIAB components.