Project NEon - The Road from Appliances to Network Elements

A new Sun Labs technical report entitled,
"Integrated Network Service Processing Using Programmable Network Devices"

by Christoph Schuba, Jason Goldschmidt, Kevin Kalajan, and Michael F. Speer

August 8, 2005 - "Why did you call it Project NEon?" was usually the first question we were asked whenever we talked about this systems research project. When we initiated Project NEon in the Network Technology Office in mid-2002, we began by asking ourselves which functions and benefits a "network element" would possess. If you take the letters NE and know your Periodic Table of the Elements, you pretty quickly arrive at the name NEon.

We arrived at the term "network element" by examining the history of how network service functions moved from having been performed within servers to being performed in special purpose network appliances. This trend could be observed starting in the early 1990's; just think of firewalls and load balancers.

The main driving force behind this trend was the need to place these functions inline into the data plane and operate these services at line rates. This technical solution has served as a sound business model for network appliance vendors throughout the last decade. However, as this approach became more prevalent and network speeds increased, the limits and disadvantages of this approach also became apparent. For example, individual appliances have problems dealing with ever-increasing bandwidth requirements. Also, stringing appliances together to build network service architectures in data centers hit architectural scaling and performance limitations. Finally, such architectures became notoriously difficult to manage.

NEon architecture components: Components comprise a control plane and a data plane, separated by a standards-compliant interface layer.

By early 2004, appliances that combine a small set of related services as an integrated solution were beginning to be offered in the marketplace. Project NEon is the first proponent of driving this approach to its logical extreme. Project NEon investigates a paradigm shift away from special purpose network appliances to an integrated way to architect, operate, and manage data plane network services. We were interested in evaluating the benefits of data flow management and enforcement inside the data center edge. We therefore investigated ways to architect data flow management over high bandwidth network connections feeding data centers, and focused our attention on handling data flows vs. individual packets.

The subsequent technical report presents not only the motivation, NEon architecture, and the project progress and scientific contributions throughout two major phases to project completion in 2004, but also devotes much emphasis to lessons learned and suggests ideas for future work.