CC-NIE Integration: Bringing SDN based Private Cloud to University Research

Stanford University, Stanford CA

Investigators

Abstract

Scientific computing groups on university campuses are often limited by insufficient access to computing cycles, storage capacity, and network bandwidth in a flexible and cost effective way. This prevents them from scaling their applications and taking on grand-challenge problems. Moreover, the current network infrastructure is built with proprietary switches, routers and with proliferating appliances such as firewalls and load balancers. As a result, the IT operation group faces increasing complexity, cost, and lack of control of the existing infrastructure. This closed infrastructure also prevents network researchers from exploring new ideas at scale in a production setting. This project leverages two disruptive technologies that the Web infrastructure and "a cloud" are exploiting: multi-tenancy virtualized clusters and Software Defined Networks (SDN). Multi-tenancy virtualized clusters enable scale-out designs with very flexible resource use and the best cost performance. SDN makes it possible to customize infrastructure and to eliminate unnecessary complexity and costs. This project builds an SDN-based Private Cloud to bring these two disruptive technologies to the campus and with them the scale, flexibility, and cost performance. With this private cloud, scientific computing groups are able to share the physical infrastructure while simultaneously customizing computing and networking for their applications. Radar remote sensing and biological computing applications are used as example scientific computing applications to demonstrate value of this infrastructure. The SDN based Private Cloud will impact various communities on a campus and will become a showcase for other campuses and thus help them exploit the latest trends in networking and computing.