SDN slicing has moved beyond initial enterprise, data center and campus-wide deployments and is now being implemented on a few service provider backbone networks, as a recent announcement from Internet2 illustrates. The organization, which operates a nationwide high-speed research and education network, said October 28 that it has successfully used slicing on that network.
I talked recently with Rob Vietzke, vice president of Network Services for Internet2, about this development and about the benefits of slicing, which uses software defined network to isolate a certain type of traffic or a certain group of uses from other traffic on a network.
This may sound a lot like a virtual private network but as Vietzke explained slicing is, “more foundational – it does what a hypervisor or virtual machine does on a server.” Another key difference is that different users can use different operating systems on different slices of the network at the port forwarding layer, he said.
Slicing is now implemented at 50 add/drop nodes on Internet2 – and authorized personnel at any of those nodes can establish a network slice between that node and any other.
Traditionally, end users have had to choose between using the general purpose Internet, buying a managed VPN from a network operator or building a dedicated network, Vietzke explained. Slicing “essentially gives you all the command and control you have in a dedicated network as part of a shared network,” said Vietzke.
In comparison with VPNs, slicing technology also is designed to offer a higher degree of certainty that users will always have access to a certain amount of bandwidth – although Vietzke noted that this capability is not currently implemented on Internet2 but is expected to be available in the future.
The idea is to, “make sure a job running on one virtual machine can’t crush another,” Vietzke said.
Potential Applications
Internet2’s slicing capability will be critical to two pioneering networking projects involving cloud computing – one at the University of Chicago and the Texas Advanced Computing Center known as Chameleon and the other spearheaded by the University of Utah known as CloudLab. Both projects target what many see as the Holy Grail of cloud computing – the ability to not only turn up computing resources on demand but also to tightly couple that capability with the ability to turn up network bandwidth on demand to support cloud connectivity.
Internet2’s slicing capability was made possible by equipment from Brocade and Juniper. The two companies collaborated closely on the project, Vietzke said, but they did not build special equipment for Internet2. That means other network operators could use the same equipment to implement slicing if they chose to do so. And according to Vietzke, both Brocade and Juniper use Open Flow software for SDN, which means their SDN capability should be interoperable with that of other manufacturers that use Open Flow for SDN.
Vietzke is aware of at least two other network operators that have implemented slicing based on Open Flow, which means that slices could be implemented that use portions of Internet2 and portions of one of those networks. Currently some manual collaboration would be required to implement such a network slice but eventually it may just as easy to establish such a connection as it is to establish a connection within one of the networks.
Edited by
Stefania Viscusi
