Tuesday, 7 May 2013

Introduction to Software Defined Networking

Networks have become an essential part of the infrastructure of institutes, organizations, businesses and homes. Depending upon the new technologies and customers’ demands, there have been remarkable changes in the present network infrastructure. This has made the structure of networking more complicated.

Moreover, the developments in the networking trends have been proved to be both blessings and curse for the networking researchers. Although their work is more pertinent, the chances of executing their ideas in the real world traffic are more distant. The reduction in real-world impact of innovation with any given network is because the tremendous installed base of equipments and protocols, and the unwillingness to test with production traffic, which have created an exceedingly big impediment to the entry for new conceptions.

There was a rare way with which a researcher could do any experimentation with his newly invented network protocols (say, new routing protocol, or an alternative to IP) in the real traffic environment. This was the key reason for most of the novelties in the area of networking remained ossified.

The demands of cloud computing and ponderous data-centers have made effective networking much more complicated. In order to cope up with these demands, operators want their networks to be more intelligent in order to be able to control and govern them in better ways. Having recognized the problems, researchers paved their feet to discover an entirely new concept- "Software Defined Networks".

These are the programmable networks consisting of programmable switches and routers that can process packets for multiple isolated experimental networks simultaneously. For example, in Global Environment for Network Innovations (GENI), it is expected that a researcher will be provided with a portion of resources across the whole network, consisting of a slice of network links, packet processing elements (e.g. switches) and end-hosts; so that researchers tackle their slices to work as they desire. A slice could expand across the backbone, industrial research labs, into college campuses, into access networks, and include wiring closets, wireless networks and sensor networks.

In order to run experiments in the campus networks, following challenges were supposed to be discussed upon -

  • How will college network administrators be comfortable putting those experimental equipments (switches, routers, access points, etc.) into their own network? 
  • How will researchers manipulate a slice of local network so as not to disrupt others who are the part of the same network? 
  • To enable experiments in the network, what functionality should the switches provide?

To overcome above challenges, commercial equipment vendors should provide an open, programmable and virtualized platform on their routers and switches, in order that researchers can deploy their experiments upon them, and network administrators can guarantee that the equipments are well-supported. But, this is very improbable to be followed by the vendors. They usually don't provide an open platform, so one should not recommend them to provide the means of virtualizing their hardware. In general, the functionalities of their equipments are confined (Packet forwarding) and internal flexibility which diverges from vendor to vendor, is hidden.

Thus, each vendor has its own API and fixed SDN functionality, which restricts the capabilities of researchers to administrate and manipulate the traffic across the equipments from different vendors.

For more information, refer following websites:


Post a Comment