PAC.C

From OpenFlow Wiki

Packet and Circuit Network Convergence with OpenFlow

Wide area networks are expensive to own from a service provider perspective and it is widely understood that much of this cost is in operational expenses. However, service providers such as AT&T and Verizon are obliged to own and operate two distinct networks for IP and Transport. These networks are typically planned, designed and managed by separate divisions within the same organization, leading to substantial management overhead, functionality/resource duplication, and increased Opex.

There have been may attempts to unify the control and management of packet-switched (IP) and circuit-switched (Transport) networks, but none have taken hold, mainly because such attempts have assumed that the control architecture of the networks cannot be changed. We believe that architectural changes will enable true network convergence. We reason that if such changes allow both types of switching technologies to be controlled in a common way, it allows the network operator maximum flexibility in using the correct mix of technologies while designing and operating their networks. To that end, we propose a converged architecture and control plane using OpenFlow.

Why Circuits?

IP routers today are connected over the wide-area by static circuits provisioned in the Transport network. Some propose connecting the packet-switched routers by direct point-to-point optical WDM links and eliminating the transport layer switching altogether. We don't believe optical circuit switching will (or should) be eliminated; on the contrary, we believe it offers significant advantages in the core of the network.

• First, optical switching is much more scalable; an optical circuit switch can switch much higher data rates, and consume much less power than an electronic packet switch. As a consequence, they are simpler, lower cost and more space efficient than an equivalent electronic packet switch.
• Second, IP and transport networks do not dynamically interact. Static circuits supporting virtual IP links, imply that packet networks cannot benefit from dynamic switching in L1 (TDM)/ L0 (wavelength/fiber) networks. But if lightpaths could dynamically created, modified and destroyed, a converged network could reap the benefits of both kinds of switching technologies - for example packet switching closer to the edge for fine grain control and statistical multiplexing benefits, and dynamic circuit switching in the core where flows are naturally aggragated and relatively smooth.

OpenFlow Unified Architecture

OpenFlow Unified Architecture

OpenFlow advocates a clean separation between control and data planes for packet and circuit networks and treats packets as flows, thereby presenting a unified architecture for both switching technologies.

OpenFlow abstracts each data-plane switch as a flow table. The control plane makes decisions on how each flow is forwarded, and then caches the decision in the data plane’s flow table for each switch along the chosen route. OpenFlow allows the definition of a flow to be any combination of L2- L4 packet headers for packet flows, as well as L0-L1 circuit parameters for circuit flows. We have made draft experimental extensions to the OpenFlow protocol to support various circuit switching technologies - time-slot, wavelength and fiber.

Unifying Abstraction

By providing a standardized open interface to the data plane for both packet and circuit switches (the OpenFlow protocol), innovation can take place at a much faster pace than today. Network operators, vendors, researchers and 3rd party developers, can all add new functionality and services to the network by creating networking applications that run on top of the network operating system, thereby helping network services evolve more rapidly, leading to a Capex and Opex efficient infrastructure.

Demos

SC09 demo

We have recently demonstrated a simple OpenFlow enabled Ethernet and TDM switched network using carrier-class hybrid packet/circuit switches from Ciena, at SuperComputing 2009 (SC09). Our main objectives were, to not only show unified control of different switching technologies, but also show the ease in creating a network application which took advantage of the unified API over the two switching technologies. The application relieves network congestion via Variable Bandwidth Packet Links. More details can be found in the OFC'10 paper below. See the demo video here. For controller & application reference code go here

In addition we have done a lab demonstration of unified control over packet and optical wavelength switches (WSS) with Fujitsu. Details are in the publication section. See the demo video here.

Publications

• Saurav Das, Guru Parulkar, Preeti Singh, Daniel Getachew, Lyndon Ong, Nick McKeown, Packet and Circuit Network Convergence with OpenFlow, Optical Fiber Conference (OFC/NFOEC'10), San Diego, March 2010. (Presentation)
• Vinesh Gudla, Saurav Das, Anujit Shastri, Guru Parulkar, Shinji Yamashita, Nick McKeown, Leonid Kazovsky, Experimental Demonstration of OpenFlow Control of Packet and Circuit Switches, Optical Fiber Conference (OFC/NFOEC'10), San Diego, March 2010. (Presentation)
• Saurav Das, Guru Parulkar, Nick McKeown, Unifying Packet and Circuit Switched Networks, Below IP Networking workshop in conjunction with Globecom'09, Hawaii, November 2009. (Presentation)

People

• Saurav Das
• Vinesh Gudla