Distributed kepler-based experiment workflow management for metabolic syndrome estimation

Abstract

Scientific workflow is evolving now more than the problem solving environment. Smart users, such as scientists or engineers, need more customized functions and information to automate their computational tasks. At the same time, due to the increase in degree of workflow complexity, workflows are deployed onto high-performance computing or Grid environments. Metabolic symptom simulation is such an example application that needs huge computing power and data storage. This thesis addresses three problems in metabolic symptom simulation experiment management using scientific workflow system. Fist, existing workflow systems are not well identified for the end-to end development life cycle of workflows. For example, design frameworks for identifying the experiment data that are ever used do not exist. Second, in the existing workflow systems, built in mechanisms do not exist that support the distributed model. The distributed model in here includes remote execution, spreading and collecting data on remote site. Therefore, users have to adapt external system for distributed execution. This approach requires re-design of existing workflows and components that result in difficulty of distributed execution. Moreover, the existing systems lack in supporting dynamic scheduling algorithm for the remote execution resource allocation to workflow component for performance enhancement that lead to difficulty of guarantee the desired Quality of Service. In this thesis, we discuss how to automate the metabolic symptom estimation experiment cycle by proposing web-based architecture and how to provide the distributed workflow execution. The main contributions include the building web based simulation data and execution management framework, improving the existing workflow management system (e.g, KEPLER) to support built-in distribute execution model and adapting a strategy decision scheme to enable distributed resources scheduling. Finally, for the evaluation of proposed syst...