Terry Slattery: High-Level Challenges to Network Automation
Terry Slattery is a veteran of the networking industry. If you’re not familiar with his track record and presence, search for “CCIE 1026” with your favorite search engine or take a look here. I’ve had the pleasure of working with Terry and it was one of the most educational periods of my networking career.
Terry listened to a recent episode of The Hedge (episodes #193 and #194 with Russ White, Tom Ammon, Chris Grundemann, and myself) and shared great feedback with us; he didn’t feel we answered the question of “Why hasn’t network automation taken off?” We decided to do a follow-up podcast based on his comments on The Hedge, which is coming soon.
But why wait until that pod gets published? Here’s a preview of our discussion with Terry and his view of higher level reasons that make automation difficult. His comments are based on his session at the Enterprise Connect 2023 conference on “Network Automation: Streamlining Processes while Avoiding Pitfalls”.
Culture Change is Hard
It’s difficult to change the networking technical team’s approach to networking. New approaches need to be developed and adopted across the entire team. Communication between different teams needs to be adopted, and oftentimes organizational structures have to change from adversarial to collaborative. The challenge of culture change is often the reason why automation projects fail (and many other types of projects as well).
Failure to Adopt a Network Automation Architecture
The industry has talked a lot about network automation and network source of truth. What we need in addition is an architecture definition that helps us understand all the functions that are needed and the interactions between them. NetworkToCode has produced such an architecture, which is immensely helpful:
https://blog.networktocode.com/post/network-automation-architecture-part-01/
Failure to Use a Network Automation Maturity Model
Network Automation maturity models can guide your next steps. Joel King at NetCraftsmen has defined a 5-phase model that allows you to judge your organization’s automation maturity so that you can determine your next steps. This ties back into culture change and automation architecture. It’s a journey and companies seldom are able to skip any of the phases. Each phase builds on the prior phases and is the foundation for the subsequent phase.
There are other similar models from various sources, such as:
https://www.itential.com/resource/guide/network-automation-orchestration-maturity-model/
https://netboxlabs.com/blog/5-levels-of-network-automation-maturity/
The existence of multiple network automation maturity models with different perspectives is an indicator that this could be a very useful concept; it deserves attention.
Lack of Commitment Across the Organization
There is often a lack of commitment across of the organization, from upper management all the way to the network team leadership. This ties directly to the culture change initiative. Is the automation initiative adequately funded? What about staff training? Can you hire a few network automation engineers (either employees or consultants)? Is the organization supportive of the work that’s needed to transition to automation? In one case a company implemented a massive automation project that cost millions of dollars. It was adopted by about 10% of the network team and soon fell into dis-use. Leadership allowed the team to continue using old, manual processes.
One concept we’ve heard from multiple sources is the concept of an "internal automation advocate" role, similar to "developer advocate" roles that we see in various companies. Some have initiated it themselves without the formal title. We think this type of support can help.
Moving Into the Details
There are many detailed issues that fall out of or contribute to the higher-level issues above. In no particular order, some of them include:
Selecting the right automation tools and coordinating on them across teams
Network engineers not being taught basic programming constructs; “'pair programming” can be a very effective approach here
Snowflake networks: it’s substantially more difficult to use automation in networks where everything is different
Lack good sandbox functionality that allows us to develop and test network automation processes
You’re probably reading this and saying “Well what about [this|that|other]?” You are correct that items are probably missing; this is not a comprehensive list by any means. We’ll be sure to hear more of them at AutoCon0.
What Are You Prepared To Do?
Terry and others are shining a bright light on the network automation challenges we’re collectively facing. At NAF, we want to bring people together to constructively make progress on culture change, automation architecture, automation maturity models, organizational alignment, and many other relevant areas of network automation; please message Chris or me on LinkedIn if you are interested in participating in these efforts - and thank you to those who have already reached out to say “count me in”.
We’re working to bring a broad set of perspectives from across the industry to collaborate: network engineers doing the work in production networks, professional services and consulting companies bringing network automation expertise, vendors and solution providers bringing new tools to the ecosystem, and more. We’re very focused on our first event in November 2023, but we are developing plans to build community and connections outside of events. If you join us in November, you’ll hear the first round of those plans.
AutoCon0: Come Join Us!
Terry can’t make it to AutoCon0, but we hope you can. Come to AutoCon0 and find the people you can collaborate with on these challenges. Register for AutoCon0 today while there is still space for you!
