Key Design Rules

The development of split architecture enables a split between functionalities that can be logically or physically grouped together. The key design rules of these splits are

  • Split / separation of forwarding and control, e.g., separation of software-centric and hardware-centric building blocks
  • Split / separation of a common control part and a network application-centric (software) part
  • Split / separation of a data forwarding functionality and a data processing functionality

 

Split of the network architecture and the SPARC proposal

The introduction of the split architecture concept into real networks would have a profound impact on the way networks are built and operated. To understand and evaluate the practical implications of the general concept, it would be beneficial to first test split architectures in research networks. Feedback from the experimental implementation will be crucial in improving the overall concept and would allow the concept to be used for further applications in networking. Initial trials focusing on the easy management and reconfiguration of research networks are currently underway at selected U.S. universities, e.g., for applications in the field of clean slate research.

However, the real benchmark for split architectures is whether or not they boost the features required for production networks. Research in this area so far has been weak. A promising new application for split architectures could be found in data centers: the networks used to interconnect a large number of servers suffer from scalability and reliability issues that could be addressed easily in the new concept. Once such issues are solved in the context of geographically confined networks, the next logical step is to investigate which advantages the split architecture might have in the context of geographically distributed carrier networks and their high requirements for quality and service.

Aside from these technical features, the split design will enable cost reductions and new market opportunities through the basic principles of modularization. This is of great importance for supporting flexible network innovations, because the development cycles of hardware and software components are extremely different and modularization supports a decoupling of the innovations from a market perspective. Moreover, modularization – the right split and/or layering approach – will enable high market volumes for specific modules (software or hardware). A well-known, yet simple example of the market benefit of such modularization is the IEEE Ethernet interface converter. In this case, a clever design of the media-independent interface enabled the mass production and clever usage of interface converters, which was the basis of the consistent price benefit of Ethernet compared to SONET/SDH modules.

Secondly, it is SPARC’s objective to develop carrier-class solutions, that is:

  • Scalability to millions of customers
  • High reliability and robustness
  • Highly automated operation
  • Improvement of investments, e.g. interoperability with existing solutions and downward compatibility