How DevOps feeds the “cattle” economy of NFV

As SDN and NFV transform the way we build and manage networks, IT is going to need to raise its game to keep up with the demands of this more dynamic and sprawling infrastructure. The methods you’d use to manage a small family farm could never scale to support an industrial dairy operation, and you can’t manage tens of thousands of virtual machines the same way you’d deal with a few hundred physical servers. In that light, DevOps is quickly becoming a crucial skill for enterprise networking. Put to work effectively, it can help you lower costs, optimise performance and improve productivity while leveraging the full agility of network virtualization.

To understand the essential link between NFV and DevOps, it’s helpful to look at what’s been happening in telecoms lately. The experiences of these organisations, while shaped largely by their specific business context, have much in common with other enterprises.

As you can imagine, the networks of telecom companies are highly hardware-intensive, encompassing appliances of all shapes and sizes, with their implications for space, power, integration and capital expense. They’re also somewhat rigid, built to support the predictable loads and needs of voice telephony. Now, as data services continue to explode in variety and volume, these companies have no choice but to go digital to improve flexibility and agility. That’s one reason NFV originated within the telecom industry—providers needed to be able to run next-generation applications on standard servers without having to wrestle with proprietary hardware technologies.

With NFV, telecom providers can build out their infrastructure far more quickly and cost-effectively, and ensure that it can meet any need with optimal efficiency. By creating an abstraction layer above the physical network, NFV makes it possible to move VMs around easily while maintaining a consistent logical address; only the mapping of its network location to the physical infrastructure changes—a task that can be handled automatically. IT can also replicate VMs to new locations while making them feel like they’re still on the same network as their original instances. This fluid flexibility and easy scalability makes it possible for companies to keep the cost of telecoms affordable while also accommodating the incredible growth in interconnected devices we’ll see as the Internet of Things takes off.

But NFV also introduces new challenges. It’s great to be able to clone VMs to add capacity for various services—a virtualized ringtone server here, a virtualized video server there. But as telecoms use this rinse-and-repeat approach to scale their business, they can easily end up with hundreds of thousands of VMs all over the world. And they’re not the familiar boxes in a datacenter anymore. As in dairy farming, you quickly go from Bessie and Clementine to a vast population of VMs known only by number. They’re easily cloned and easily replaceable, and there are far too many of them to get to know on a personal basis.

This is where DevOps comes into play. Instead of having developers crank out an endless assembly line of VMs for operations to deal with on their own, developers collaborate with operations to deploy and support applications. These applications are built from smaller, task-specific micro-service components that can be replicated easily, and orchestrated according to the needs of the specific service. That provides a much more efficient and flexible way to build and adapt services, and even alter the purpose of an entire datacenter at a moment’s notice, making what used to be difficult and costly far more manageable.

DevOps changes the way you think about system admin. Once you build and debug a configuration template, you can stamp it out over and over again to scale—or have code run that task for you. This involves using an engineering mindset in terms of how to manage and version this code to make sure that any given engineer can come in and take over with a complete understanding of how things work—a departure from the traditional approach to scripts as one-off hacks—but that’s an evolution every organisation should be making. We have to treat everything we do as the rule, not the exception, to ensure rigor and consistency and avoid mistakes.

By using DevOps as a tool to manage their networks, telecom providers can see their cost of providing services start to level off instead of increasing geometrically. The amount of work required to manage the infrastructure becomes more finite and limited, lowering the personnel costs that contribute such a large proportion of operating expense. Meanwhile, as NFV expands your own personal network management capacity logarithmically, your value as an individual goes up. You can only keep so many physical machines in your head at the same time, and you can touch only so many servers in a day, but NFV lets you apply your talents to a far larger network.

As is so often the case with DevOps, the biggest challenge to address is education. It’s hard to find enough people with both DevOps and virtual networking skill sets, limiting the ability of telecoms and other enterprises to adopt these new deployment methods. That’s one reason I usually tell customers to expect this to be a two to five year journey, not an overnight transformation. On the other hand, for those who take the initiative, there are ample resources to learn these skills, from free programming courses online to numerous books. All the scripting languages are publicly documented, and if you have system admin skills or know how to automate Linux and Windows, you’re already ahead of the game.

And the effort is well worth it. Once you figure out the right way to solve something, you want people to be able to repeat it over and over again. That’s how you scale out an IT organisation—and when you provide a playbook for people to follow, mistakes plummet. Automation helps you lower operating costs, optimise performance and improve productivity. Those are goals any IT organisation can embrace—in telecoms or any other industry.

Steve Shah, VP of Product Management, Delivery Networks at Citrix

Image Credit: Shutterstock/Sergey Nivens