PNG-24 image with transparency
PNG-24 image with transparency

Clean Slate Research Projects

1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10

9.) Fast Dynamic Optical Light Paths for the Internet Core

Started: September 2006

The core of the Internet has evolved over the last 20 years. In the 80's and most of the 90's, IP packet traffic was adapted and carried in TDM circuits optimized for carrying telephony voice traffic. This trend was further continued with the addidtion of the WDM layer beneath the TDM layer, in effect increasing the capacity of the backbone by connecting the SONET/SDH multiplexers and crossconnects with multiple wavelengths. However, in the last few years, IP backbone's have been designed to be directly supported on the DWDM layer. This so called IP over WDM architecture (sometimes termed IP over optical), incorporates SONET/SDH (or OTN) framing for IP packets, but no SONET switching. The underlying DWDM transport network, supports both TDM traffic as well as IP traffic, the latter being the dominant form of traffic since the turn of the century.

The DWDM transport network

The DWDM transport network provides the long haul interconnection between large backbone routing centers known as Points-of-Presence (PoP). A typical PoP is consists of several smaller access routers, which aggregate data delivered to them from customers via metropolitan transport networks. These access routers are then linked to two or more large backbone routers, which in turn aggergate and direct the traffic over the core via the hardwired lambdas in the DWDM transport network.

Why Do We Need To Rethink Internet Backbone Design?

Driven by an increase in rich media applications (eg. video), service convergence, and broadband access deployment, global IP traffic is showing explosive growth. Furthermore, IP data network traffic is highly dynamic with constantly changing demands and substantially different characteristics when compared to telephony voice. IP backbones are designed to have the core routers directly connected via static circuits (wavelengths). This assumption of static links results in a need for the links to be grossly overprovisioned (2X to 4X) to deal with peak traffic uncertainties and link failures. As IP traffic grows, such a network design simply implies two things - bigger links and bigger routers - neither of which are cost effectively scalable. We further note, that the large costly backbone routers, are often unrealible and present a serious problem with power consumption and heat dissipation.

.. & How Can Optics Help?

We start with a Clean Slate assumption of removing the static links by introducing photonic switching in the transport layer. Photonic switches offer ten times the capacity density (in Gb/s/cubic meter), consume a tenth of the power and cost less than a tenth (per Gb/s) when compared to the backbone routers. Further, we propose a flexible IP/transport architecture optimized for carrying IP traffic, where access routers are directly connected via a switched, shared resource of meshed photonic switches, to other access routers. Finally, we propose that the IP routers dynamically control the setup and teardown of the optical circuits that connect them. In this project, we seek to answer several questions regarding the router functionality, and the IP and transport network architectures, in our goal of achieving a simple, elegant, flexible and cost-effective future Internet core.


Nick McKeown
(Faculty) Faculty Director, Clean Slate Program

Leonid Kazovsky Leonid Kazovsky
(Faculty) Professor of Electrical Engineering

Saurav Das Saurav Das


This research is funded by the Clean Slate Project.

We will post papers/progress reports on this page.


1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10