Feb 26, 2014
Software Defined Networks (SDNs) have obviously been a hot topic on this site and many others for the past few years. Despite the media hype, SDNs remain somewhat of an enigma for many enterprise network mangers. The concepts of automation, programmability, and centralized control make sense to most network professionals I have talked to. However, where to apply the technology and what problem to solve isn’t really that obvious. From a utopian perspective, one would want to apply the principles of SDNs to the entire network, holistically, and start enjoying the benefits right away. In practical terms, though, most organizations just don’t have the tolerance for change and risk to do this. After all, the network is the lifeline of most organizations. Without the network, organizations would no have ability to pass information from person to person or location to location, effectively crippling the business.
So why not apply SDNs to where the biggest source of pain is? Although most of the SDNs vendors have targeted the data center with solutions, the wide area network (WAN) is where more organizations have challenges running the network. The current hub and spoke model has been in place as long as I can remember. Sure, we’ve been talking about moving away from this architecture for years now, but it seems like it never happens. Hub and spoke was sufficient when client server computing was the norm, although having all the Internet traffic “trombone” up and down the WAN wasn’t ideal. However, this was the minority of traffic so the model worked.
Today though, traffic patterns have changed significantly. Cloud and mobile computing have made Internet traffic mission-critical, and have increased the volume of Internet traffic by at least 5x over the past 5 years. Additionally, WebRTC, video, Microsoft Lync, and other collaborative tools have increased the amount of peer-to-peer traffic by orders of magnitude.
The proposed answer to many of these WAN challenges has been to increase the level of ‘mesh-ness’ of the network. The degree of meshing required really depends on the organization and the type of applications being run. As good as this sounds, in practicality, it’s very difficult to run a fully or even partially meshed network. Troubleshooting can be difficult and configuration changes can be a nightmare with the traditional box-by-box management model of running a network. Additionally, determining where to create links, how to manage them, and when to terminate them all add to the complexity.
SDNs can help organizations increase the level of meshing without having to add to the complexity. In fact, if done correctly, the creation of new paths should be automated based on application needs. For example, let’s say an organization was running video conferencing. The video application should be able to talk to an SDN controller and then tell the controller to create a link between the two locations engaging in a video call. Once the video session ends, the controller can send commands to the network to terminate the link.
In a sense, SDN enables organizations to always have the exact level of meshing configured to optimize application performance — without the configuration headaches. The ability to orchestrate network changes with application requirements combined with the automation capabilities of SDNs makes it an ideal solution for WAN management.
So for those looking to figure out what to do with SDNs, many look past the data center and focus on the WAN where many of today’s network management challenges reside.
Image credit: Hugh Nelson (500px.com)
Zeus Kerravala is the founder and principal analyst with ZK Research. He provides a mix of tactical advice to help his clients in the current business climate and with long term strategic advice. Kerravala provides research and advice to the following constituents: End user IT and network managers, vendors of IT hardware, software and services and the financial community looking to invest in the companies that he covers.