Test and development
The first benefit of SDN for developing the network was that it worked really well in a multichassis environment. This can be difficult for vendors, getting an application to work consistently across a range of devices.
With the control sitting outside, we could make multiple devices function as one logical device. Also, rather than being limited to the relatively small processing power within in a router, we could do much more intelligent management with the power of high performance servers hosting the local controller. We also used existing distributed system software to support controller failover. Then, with SDN now in place, it enabled ongoing updating and addition of new traffic engineering and security applications - Fig 3
When it came to testing, SDN offered a load of benefits, mostly around simulation. The OpenFlow agent, using V-Switch, is already a simulated network device and it is easy to run a mass of v-switches as a very large network. Even when the hardware is not yet available, control software can be tested on a very large scale, and new configurations or control tools can be trialed in advance without compromising a working network.
An extension of this that we found very useful was to combine real and simulated network elements in the test, verifying the behavior with real devices while scaling the test to a greater network. One possible extension of this, yet to be tried out, would be to create an entire synchronised mirror network, subjected to the same real-world event stream but without executive action on the actual data flow - so that changes could be tested in real conditions but without putting real data at risk.
I outlined these powerful test opportunities made easy by OpenFlow at a recent presentation and one service provider came up to me and said "you've got it exactly right" - meaning however much work goes into the design of big networks, testing is the real key to success. Tests like those described could also be generated using traditional routers, but they become a whole lot easier with OpenFlow, and that is very important. Network testing is a vital but Cinderella task, and the more it is made easy and quick, the greater chance it won't be passed over.
[Turn to next page]
Sometimes, the power of the mainframe is the most cost effective answer. Computing's Peter Gothard puts Computing's readers' questions on the future of the mainframe to IBM's Z13 expert Steven Dickens.
This Dummies white paper will help you better understand business process management (BPM)