A Convergence Architecture for GRID Computing and Programmable Networks

Abstract

Abstract. GRID computing in the Internet faces two fundamental challenges. First the development of an appropriate middleware for provision of computing services in a scalable and secure manner. Second the distributed, scalable and fast connection of GRID processing components to the network. The latter is important to enable fast data exchange among the distributed components of virtualized super-computers. In this work, we focus on the connection aspects of GRID architectures and advocate to enhance edge-routers with clusters of high performance computing machines. We propose enhancing available GRID resources to a componentbased programmable node. Our approach delivers three advantages: Use of the same hardware/software for GRID computing and Programmable Networks. Support of GRID computing through-then available-programmable services, like data transport and distribution in the GRID through Programmable Networks. Finally we see GRID computing as a leverage for the future deployment of Programmable Networks technology.