The Packet Pushers have been covering software-defined networking (SDN) for years now. In its early days, SDN was hyped as a game-changing, transformative technology that would upend both the networking industry and the way networks were designed and operated. SDN was going to change everything.
Time passed, and vendors battled over products and standards, new startups emerged with new ideas, and jaded network engineers decided SDN stood for “Still Does Nothing.”
So where are we today? In this Packet Pushers podcast, recorded live at Interop Las Vegas, Greg Ferro and Ethan Banks, check in on the state of SDN in 2016 to look at what’s changed, what hasn’t, whether SDN means anything any more, and why you should start thinking about use cases instead of worrying about whether something is software-defined.

Sponsor: GNS3
The GNS3 Academy is offering the Practical SDN and OpenFlow Fundamentals course from David Bombal, and you can get 50% off. If you re looking for a practical, hands-on, real-world approach to learning and implementing SDN and Openflow, look no further. The first 50 listeners to register receive the course for 50% off. Go to tinyurl.com/gns3sdn OR go to academy.gns3.com and use coupon: PACKETPUSHERS to get your discount.
Show Notes:
Section 1 – Forget About SDN–Let s Talk About Use Cases
- SDN isn t a thing you buy, just like you don t buy a routing protocol as such
- We know from talking to vendors that they don t like talking about SDN anymore. The term is meaningless
- You don t want to SDN-ify your network. You have a problem you want to solve. SDN might be part of that solution. Or it might not be. It doesn t really matter…
Section 2 – The Use Cases Drive The Software In Software Defined Networking
- Analytics
- Digging deep into networking data, finding patterns, making correlations
- Software: Kentik & Deepfield
- Machine learning
- It’s the latest marketing term
- It s really a nascent field with limited usefulness right now
- Big math predicts what s next based on what s past
- Networking applications? We ll see. Could be dynamic path adjustments based on historical traffic patterns, that sort of thing
- If your marketers are all excited, make sure you get right down in the weeds with them about it. What are the machines learning, exactly? And what do they do with the information?
- Policy management
- Stop managing individual devices
- Manage the network of devices centrally
- Distribute centrally, enforce globally
- Software: SD-WAN is a good example of this
- Software: Cisco ACI, VMware NSX, Nuage Networks, but from a DC point of view
- Configuration automation
- Commonly called network automation, but configuration is what we re getting at here
- We still have a need to configure individual boxes, but we want to do it better
- Less human error
- Higher speed
- Integrate with larger processes (hyperconvergence)
- Software that can tell the actual state of the network vs. the desired state, bring desired state about. This is not a system to dumbly push changes to devices
- Network modeling
- YANG
- IETF & OpenConfig
- APIs generally
- Software: Ansible, Veriflow, Glue Networks