OpenFlow3

Finding one’s way around the increasingly populous Software Defined Networking (SDN) ecosystem is initially quite easy.  The most immediate and informative non-commercial links that I found are:

This ecosystem is very active, and there are clearly many very smart people investing time and energy in driving new wave in networking.  Interestingly, quite a bit of the content that I encountered in the various articles and whitepapers is dedicated to the justifying SDN versus the details of how it works.  Suffice to say, the case for SDN is readily made many times over in the links above.

But in a nutshell, the explosion of bandwidth, the network impacts of “big data”, the proliferation of devices with access to many different network resources – all with different mixes of security / privacy / multi-tenant requirements – and the need to readily manage network configuration, access, security and control in the face of all of these new pressures on IT staff make the case for SDN readily.  However, what really amplifies the need for SDN is the revolution taking place in virtualization.

The dynamic nature of virtualization in bringing up and down capacity and server bandwidth as compute needs change has tremendous economic value.  But, efficiently managing the underlying networks to support this dynamic behavior puts a tremendous – perhaps untenable – burden on IT staffs.  The longer term network designs and configurations that used to make sense no longer make sense as network loads become unpredictable.  For example, the extensive changes that virtualization and big data transactions can cause to “east-west” traffic makes will require extensive re-balancing, tuning, and network reconfiguration an almost constant activity, particularly in larger data centers.  The SDN argument is that the control plane needs to be abstracted from the many disparate switches and routers (the data plane) resulting in a massive simplification, creating at the same time, the opportunity for IT staff and application providers to innovate independent of transport design and capability.  It is a great argument, and one supported by all of the demands we see in these new networks.

One of the very interesting technical links I came across was on the openflow.org site at http://www.openflow.org/wk/index.php/OpenFlow_Tutorial – this link walks through the installation of openflow (several alternative instances such as Floodlight, POX, Trema) and setting up and testing an openflow network in a virtual environment.  This was very informative, and helped to bring home how the system actually works.

On the deployment side, the ecosystem is more difficult to sort through.  Large innovators like Google are already using this technology.  And, while the open source community, those with clear vested interest (such as VMware), and various invested start-ups are all over this, the level of engagement by the large network equipment providers is mixed.  Doubtlessly, the investment to adopt this is high, and the potential strategic impacts to incumbents are equally high.  I will continue to report on this as I sort through the commercial side more.

Nevertheless, I found some very useful commercial and “semi-commercial” (open source, but a commercial business) information in these links:

My next area of focus on SDN will be to attempt to better sort out what is being deployed now, and of that, what is well suited for enterprise, larger data center, or service provider environments.