As the IP networking industry matures, we experience a paradigm shift in the implementation of the IP routers. The new generation of applications for IP networks requires considerably shorter product development cycles, due to the renewed interest on extensibility and short turnaround times for the introduction of new products.
This has led to the use of off-the-self components in router design. An increasing amount of commercial routers implement their control plane using commodity motherboard designs and processors that are used in personal computers along with modified commodity operating systems like Linux and the various free versions of BSDUnix.
Thus, we believe than in the near future routers should and eventually will be built with commodity controller hardware, will use third party Network Processors for their forwarding path processing and open software for the operating system and routing protocols. Among the multiple research issues involved in the design of such a router we focus on the programmability issues.
The use of programmable forwarding plane components results in a hierarchy of processing elements within the router At the top level of the hierarchy we have the general-purpose processor implementing the control plane while at the bottom of the hierarch y we have the Network Processor.
The existence of the semultiple levels where operations can be implemented, opens up many possibilities for re-thinking the conventional wisdom on how to implement networking services and creates a variety of research issues.
We plan to address the problem at two levels:
- control plane, that runs at the control processor and has access to more resources but is far from the data flow and,
- forwarding plane, which runs in specialized programmable hardware, and has less resources available but it is closer to the data.
Overall, our work in the control plane will be more development oriented, while the forwarding plane work will be mostly geared towards research.
Call for proposal
See other projects for this call