The nature of networks is changing as traffic volumes continue to increase and the quality of application experience becomes the driving factor in technology choices and designs.

While the adoption of software-defined network (SDN) technology has had many false starts and dead ends the move towards the latest incarnation of SDN solutions is accelerating.

Just a few years ago it seemed like it was only a matter of time before the first SDN solutions would become a part of the networking mainstream.

Then the move to cloud and the popularity of distributed application architectures started to alleviate many of the challenges in highly centralised and traditional network infrastructure that SDN solutions were designed to solve.

While the precise technologies, acronyms and methods involved may have changed the trend towards highly automated, programmable and even autonomous network solutions has continued.

Infrastructure Virtualisation platforms, as well as Cloud Providers, have invested heavily in expanding the type and number of network services their platforms support while exposing them through their existing automation solutions and Application Programming Interfaces (API).

The DevOps movement has also done a lot to promote highly automated, software-defined infrastructure as a way of creating better business outcomes and has helped educate IT Operations organisations on the types of processes, tools and patterns that are necessary to operate IT infrastructure in this way.

IT Organizations have known for a long time that most outages and issues occur during scheduled change windows and the deployment of a new version of an Application is something that traditionally required a very long window of downtime to execute.

Engineers would work long hours into the night over the weekend or during holiday periods frantically trying to get the new version of the Application working before the rest of the organisation and their users came back into work.

This is exactly the type of situation that DevOps teams avoid and the way they deploy is one of the most crucial techniques.

Instead of doing in-place upgrades they deploy the new version of the Application in parallel to the existing version.

They then use the load balancer to direct users to the new version knowing that they can easily and instantly reverse this decision if things go wrong.

Blue/Green and Canary deployments are good examples of the way load balancers have moved from being a relatively passive part of the network infrastructure to a key player in high-velocity Application development.

In the case of a Blue/Green Deployment (sometimes called ‘Red/Black') the load balancer is configured to send all Users to the new version of the Application.

If any problems occur the load balancer configuration is reversed and all Users return to the older, working version of the Application.

With a Canary deployment, only a small percentage of Users are redirected to the newer version of the Application, maybe 5-10%. If things appear to be working as expected then a larger percentage will be moved over, perhaps 25-40%.

This incremental cutover continues until the Team is confident that the new Application is stable and performing as expected.

If at any stage issues are discovered then Users are redirected back to the previous version.

Both these techniques are simple and dramatically reduce the cost and risk of deployments and upgrades.

When the type of automated deployment, configuration, and monitoring APIs are connected to Continuous Integration and Deployment (CI/CD) pipelines they become even more powerful and many deployments can occur in real-time with little or no human intervention.

At Kemp, we still regularly see Customers who haven't taken full advantage of these types of patterns and practices and escalate to us to help them recover from a bad deployment.

As more IT Operations teams gain confidence in these new methodologies and leverage API-driven management we will see better business outcomes.

Programmable network technologies like load balancers are an essential tool for any organisation striving to create high-velocity Application Teams.

SDN is dead, long live SDN!