Membrane computing inspired approach for executing scientific workflow in the cloud

Abstract

The continuous expansion and appreciation of the service oriented architecture is due to the standards of loose-coupling and platform independence. Service-Oriented Architecture is the most commonly and effectively realized through web services, and their temporal collaboration commonly referred to as web service composition. In the present scenario, the most popular variant of composition is service orchestration. Orchestration is achieved through a centralized ‘heavyweight’ engine, the orchestrating agent, that makes the deployment configuration a massive ‘choke-point’. The issue achieves significance when data and compute intensive scientific applications rely on such a centralized scheme. Lately, a lot of research efforts are put in to deploy a scientific application on the cloud, thereby provisioning resources elastically at runtime. In this paper, we aim at eliminating this central ‘choke’ point by presenting a model inspired from ‘Membrane Computing’ that executes a scientific workflow in a decentralized manner. The benefit of this paradigm comes from the natural process of autonomy, where each cell provision resources and execute process-steps on its own. The approach is devised keeping in mind, the feasibility of deployment on a cloud based infrastructure. To validate the model, a prototype is developed and real scientific workflows are executed in-house (with-in the Intranet). Moreover, the entire prototype is also deployed on a virtualized platform with software defined networking, thereby studying the effects of a low bandwidth environment, and dynamic provisioning of resources. © Springer International Publishing Switzerland 2014.