The most important aspect of automation is the abstraction of the services being automated. This hides the detailed information under the hood, leaving only the essential details in the abstracted service. These are often minimal, such as a unique name identifying the service. The automation includes logic that generates the required details from the abstract model and stores them.

Building abstractions almost always requires network environment-specific design and implementation.

Example:

An example of an abstracted service is a Virtual Network. It could consist of an MPLS network VRF, associated firewall connections, and MPLS sites. Such a setup can be created with very minimal input—often, just a name is sufficient. All the necessary detailed information, such as route-target values and IP addresses, is allocated by the automation system. These details should not need to be considered in daily operations, such as adding or removing virtual networks.

It is essential to construct a detailed target state of the network beneath the abstraction layer. The target state includes following elements:

• Network devices and their ports

• Connections between network devices

• A database of IP addresses at the subnet level

• IP addresses of individual ports on specific devices

• VLAN IDs

• AS numbers

• Etc.

• How all these are related to each other

The purpose of the intended state is to enable its complete delivery to devices at a later stage. Typically, it is built in a dedicated system known as a Single Source of Truth. The intended state is vendor agnostic.

The detailed intended state of the network is translated into a format supported by each device vendor in use.

There are differences between vendors in how well they support automation. Some vendors provide excellent API interfaces and libraries for automation, while for others, more custom development work is required.

The final step in automation is to deploy the generated configurations to the devices. This layer includes data validation and error handling.

Additionally, various pre-checks can be performed for the planned network changes. The configuration delivery pipeline can be integrated with test in simulated environment to ensure the changes do not cause unexpected disruptions when deployed into production.

There are also vendor-specific differences in this area. Some vendors provide solutions that handle a significant portion of these tasks.