Software Defined Networking Use Cases

by ONUG Board of Directors

 

As SDN makes its way from concept to product to trial, most IT architects are building a long list of demands, but seeing relatively few practical solutions. To date, the primary real use cases have come from Internet behemoths such as Google, Amazon, and Facebook. And so, on one hand, architects can take heart that if SDN concepts and implementations work for these hyper-scale Internet companies, chances are they’ll do just fine for those with networks of a more modest scale. On the other hand, hearing Google or Amazon success stories, and realizing that there’s a ton of engineering power behind the relatively few applications that those companies offer doesn’t translate well into the budget-conscious reality of most IT planners who must field thousands of apps with less engineering talent. Add to that forward looking challenges which will greatly increase the size and scope of networks, and it becomes clear that enterprise users will have a completely different but equally demanding set of uses.

The Open Networking User Group set out last fall as part of its two day event hosted by JPMorgan to lay out some use cases and use priorities for both its members, and for vendors and integrators.  Prior to the conference, ONUG’s board identified a total of eight use cases and during the conference attendees voted three of those cases as the priority for the next year.

As developed by ONUG, these use cases are very high level, concentrating on the outcomes and the basic technologies required for an open implementation. As such, they provide the starting point for RFIs including the open technologies to look for.

The three cases for top priority are: Integrating layer 4-7 services to eliminate appliances, taking an open approach to virtual networks/overlays, and leveraging branch WAN-SDN to reduce appliance clutter and improve WAN service value. What’s common to these three cases is a heavy reliance on fairly complex software without a necessity for dedicated custom hardware. However, as the market exists today, all of these services are delivered on custom hardware, greatly increasing both operation and capital acquisition costs.

One of the first benefits of open networking systems will be that these software intensive functions which have resided in appliances containing both control and data planes for a given application can be broken apart so that data plane functionality remains in economical and open networking hardware, while complex software which perform functions ranging from security operations to application optimization functions to data caching can all be handled on typical commodity servers.

Operations can be improved by allowing these functions to be run as just another set of virtual machines, managed like every other virtual machine within or outside of the data center. This is a goal keenly desired by ONUG members who want to move their network engineers from their current path, where each manages a hundred or so networking devices to the path of their server managing colleagues who routinely manage thousands of devices and virtual machines.

As ONUG’s members look down the road toward building networks that support everything from big data systems to millions of mobile users to the potential of increasing network nodes by a factor of ten to a hundred by incorporating Internet of things technology, they realize that physical appliance proliferation as means of growing the network simply can’t continue. It’s this future-looking view that has pushed these use cases to the forefront.

The ONUG Board will meet shortly for a day long summit to review the state of the use cases. At this year’s spring ONUG conference, attendees will again vote on the use cases most important to them for the upcoming year, as well as evaluate the progress on the priority uses cases for this year. We expect that the list will be at least somewhat similar to this year’s list – as these challenges are far from solved by the current state of the industry – along with a few possible new ones. We hope you’ll join us to cast your vote.


Leave a Reply

Your email address will not be published.

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>