Network Automation Concepts

This section introduces data models and helps you answer these questions:

• How should I structure the data I need to describe devices in my network?
• Should I describe devices or should I focus on network topology and services?
• Can I use device-level data to create device configurations while presenting high-level data models to the operators?
• How could I transform abstracted (high-level) network- and services data models into device-level data models?
• How could I integrate data model transformation into my automation workflow?

After you built your data model, you have to create data structures (device, link, network, service, or customer data) in a format readable by humans and machines, describe the data model for documentation and validation purposes, and store the data somewhere.

This section dives deep into presentation formats (XML, JSON, YAML), data description languages (including YANG), and data stores, from text files and Excel to NoSQL databases.

Testing and input validation are extremely important components of any relaible software solution, and you should not skip them while developing a network automation solution.

This section describes the software testing concepts, unit tests, and integration/system tests. Input validation and integration/deployment pipelines will be covered in an upcoming live session in early 2023.

Once you have a reliable test suite, you can start implementing continuous integration (automated merging of changes to the baseline code branch), delivery (having a deliverable environment at all times) and deployment (automated deployment of the deliverable environment).

This section describes CI/CD principles, their applicability in network automation, and the hidden gotchas of trying to validate network configurations before deploying them.

After mastering the data model-, model transformation-, and data store concepts, you're ready for one of the most important network automation topics: single source of truth.

One of the first steps on anyone's automation journey should be tight control of device configurations using a version control system. This section describes how you can use Git and GitLab/GitHub to track changes to device configurations, correlate changes to tickets or business requirements, implement review and approval workflow, and finally use Git as the single source of (configuration) truth.

Network Infrastructure as Code (NIAC) is another overhyped buzzword. In this section we'll try to define what NIAC is, why you might care about it, and how to implement simple NIAC in networks that use decent devices with configuration replace functionality.

The second half of the presentation introduces advanced challenges like abstraction layers, dealing with numerous devices and byzantine failures, and using software development methodologies in network automation.

Most of the intent-based systems are nothing more than a fancy orchestration system with an abstraction layer. This section describes the many levels of abstraction you can implement in such a system.

After automating device configurations and service provisioning you might start tackling the holy grail of network automation: changing the network behavior based on real-time external event.

In his March 2018 presentation, David Gee described the fundamentals of Event-Driven Automation (EDA), including:

• Why would you want event-driven automation and what are its pitfalls?
• What exactly is an event and what's the difference between signals and events?
• How would the architecture of an event-driven solution look like?
• Why do we need event normalization and correlation?

He concluded with an overview of open-source and commercial tools you could use when building an event-driven solution and demonstrated the concepts with two simple examples using StackStorm and Salt.