Recently, I went on some customer sales calls where all three companies operate hyper-scale data centers as the foundation of their business model. One provided cloud services, another was a portal while the third was in financial services.

Despite meeting people at different departments and levels within these respective organizations, a couple of common themes surfaced at all three companies, even with vastly different applications driving their businesses.  It helped me realize that we have some challenges that must be met before software defined networking (SDN) can gain wide adoption.

First, by visiting with the respective architecture teams, I learned that  represents the core technology and operational language throughout these organizations.  These teams were all quite adroit at leveraging Linux to make programmatic changes which the business required in just hours or days.

By then meeting with the respective network teams, it was apparent that “coding up a switch with CLI” was the common operational mindset or language because if anything more significant was required, you were at the mercy of your friendly infrastructure vendor because there is no community like Linux and therefore no other way to get what you need. These networks were still driven mostly by manual changes using CLI.  Unfortunately, programmatic change could only happen if it was driven through new projects, like 1-10 Gig I/O upgrades or changing your topology to be more Layer 3-centric. That meant you probably had to deploy BGP to the rack edge. These types of projects take several months and often cause a fork lift upgrade of old hardware.

Clearly, the network teams exposed us to what I’ll call “language” or “cultural” barriers that will need to be bridged for organizations to fully embrace SDN.  SDN should help these issues by driving an overall orchestration between the network and application needs, where server/compute, network and even storage maintain a common framework for change and programmability.

If SDN is to build these bridges, it needs to be universally leveraged by data center personnel not just the developer guys, not the networks guys—everyone. Common scripting languages, such as Python have the potential to be widely utilized in order to program devices. When will this happen? No one knows, but the sooner it happens, the faster SDN will be that bridge for those barriers.

We will have some forthcoming blogs on the topic of where Python based programming makes sense.