Progressing Network Automation & Why AutoCon0 Matters

Can you talk about your thoughts on the current state of network automation generally? Successes, challenges, current impact?

Network automation continues to progress along a path that finds incremental wins within a variety of different teams. However, as more and more organizations embark on their network automation journeys, there tends to be a constant start-stop motion due to a variety of external factors.

Current firefighting, DIY vs COTS (commercial off-the-shelf) debates, common network models, and the debate over centralized vs decentralized source of truth are all impediments that continue to plague organizations’ ability to holistically embrace organization wide strategies around network automation.

There is also still a very significant faction of the industry that questions the viability of network automation’s impact to the overall efficiency of infrastructure operations. From engineers who would rather embrace the CLI, to an inability to move off of bespoke device configurations, network automation still has some people to win over.

And what about the teams that are finding success? Or is everyone getting stuck?

No, not at all. All of that said, there has definitely been significant traction with teams that have committed to their own automation journeys, instead of getting caught up in endless debates over process. By staying focused on achievable use cases and avoiding “boil the ocean” strategies, organizations are gaining newfound benefits by leveraging network automation.

Of course, there is still much to do. By many metrics, network automation is still very much in the nascent stages of the technology adoption lifecycle. I suspect we will continue to see a fair amount of innovation coming from all ends of the spectrum.

Why do you think we haven’t yet seen full adoption of network automation? Is there an easy answer? Technical reasons, cultural, something else?

There is no one thing that is holding the industry back from full adoption of network automation. But, to drill down, I don’t believe that the problem is a technical one, especially not anymore. It definitely seems to have shifted to more of a cultural challenge.

In my experience, there is still too much emphasis placed on the approach to network automation, as opposed to the practice of network automation. Getting into never-ending academic squabbles about sources of truth, or DIY vs COTS, or anything else doesn’t bring you any closer to full-fledged adoption. And this is happening across the industry.

Instead of getting caught up in perfecting an approach, I think we as an industry need to stay closer to the use cases and understand what does work and what doesn’t. For those things that don’t work, why not? Is it a skills gap? An approach gap? We have drifted a little too far away from the use cases and found ourselves caught up in the implementations.

And how do you see network automation evolving going forward?

If only I had that magic crystal ball to show me the way.

Personally, I believe there will continue to be more and more DIY, or FOSS, network automation projects that will be announced. The industry still seems mostly jaded towards COTS implementations and vendor solutions.

There will also be a renewed (or continued, depending on your perspective) push towards integrating configuration with network state. Pushing configuration changes to a network device isn’t hard; however, validating that the network is operating as intended still feels a little bit out of our grasp. I believe there will be much more work in this area.

Finally, I believe that cloud will continue to create more and more headwinds with regards to the advancement of network automation. Application development teams continue to find new and creative ways to circumvent the sluggish pace of current self-service networking. The rise of Network as a Service options is a real deterrent against building large-scale infrastructure for organizations that are fully committed to the cloud.

So looking at these possible evolutions, how might network practitioners, engineers, and architects want to think about network automation to help advance their approach?

Like I’ve said, I think the critical piece is simply focusing on the use case and what exactly you are trying to solve. I have witnessed far too many teams start with very good intentions, and even a nicely defined, achievable scope around network automation — only to be sidetracked by implementation details and debates that are mostly meaningless in the long run.

Everything doesn’t have to be perfect up front to start automating. That’s the part many teams and practitioners seem to overlook. If you leverage a feedback-based way of doing things, consistently reevaluating and reassessing progress, you can make the necessary course corrections as you go. I believe that will unlock the key to advancing automation efforts.