Network Virtualization

Description | Publications | People | Collaborators | Funding


Network virtualization provides a powerful way to run multiple networks, each customized to a specific purpose, at the same time over a shared substrate. Our research on network virtualization focuses on two main scenarios. First, we consider the role of virtualization in running multiple experiments simultaneously in a shared experimental facility. For example, the NSF
GENI initiative focuses on the design and deployment of a shared, wide-area experimental facility to support a wide range of research in networking and distributed systems. The VINI project (radio interview) is a step in that direction, supporting experimentation with new routing, forwarding, and addressing schemes on a shared facility built on top on general-purpose processors. Second, we consider the role of virtualization to support multiple architectures simultaneously as a long-term solution for the future Internet. The Cabo project explores the benefits of running customized architectures, as well as how a virtualized system enables an economic refactoring of a future Internet into infrastructure providers (that own and operate the equipment) and service providers (who lease virtual components and offer end-to-end services to users). All three projects grapple with the technical challenges of providing a virtualized, programmable substrate that operates at high speed; the Cabo project must address the additional challenges of building a substrate that can operate without any dependency on the existing Internet.


Using virtualization to improve network management

Network architectures based on virtualization and programmability

Building and running a virtualized, programmable network infrastructure

GENI: Global Environment for Network Innovations (web site)




These projects are funded by grants from the National Science Foundation (NSF). The initial work on VINI was partially funded by an HSARPA grant. In addition, the Abiliene Internet2 and National Lambda Rail backbones have generously provided the VINI nodes with co-location, bandwidth, and hands-and-eyes support.