Automation in Network Management
Networks used to be configured one device at a time, by hand, over a console. That does not scale, and it is where most outages come from. This lesson explains why the industry moved to automation — and the two ideas, configuration drift and idempotency, that underpin every automation tool you will meet this week.
Why automate the network?
For 30 years the standard way to configure a Cisco device was to open a terminal, connect over the console or SSH, and type commands into the command-line interface (CLI). Each command changed the running configuration of exactly one device. That works well for one router, and breaks down the moment you have hundreds or thousands of devices.
Network automation is the practice of driving devices with programs and data instead of by hand. It replaces a sequence of human keystrokes with a repeatable, testable, version-controllable process. The benefits are usually summarised in four points.
The key insight is that human typing is the risk. Every manual command is a chance to make a mistake — a wrong wildcard mask, a mistyped VLAN, a shut interface that should be up. Research on outages repeatedly finds human error during change as the leading cause. Automation does not eliminate human judgement; it eliminates human typing, and typing is where most mistakes happen.
The hidden cost of manual configuration
Consider a simple change: add a new VLAN to a campus of 200 access switches. Manually, an engineer must connect to each switch, enter configuration mode, type the VLAN, verify it, save it, and move on. That is 200 repetitions of the same six steps.
- One switch is skipped because the engineer’s list is stale — a partial change.
- One switch is reached in a different software version where a keyword differs.
- A command is mistyped on one device and nobody notices until users on that floor complain.
- The change is never documented, so the next engineer cannot tell what happened.
Automation attacks every one of these failure modes. A script holds the authoritative list of devices, applies the same change to all of them, records what it did, and can be run again safely if it is interrupted. The list of devices and the change itself become data — stored in a file, reviewed like code, and versioned with Git.
Configuration drift
A device configured by hand, then by another engineer, then by a third, slowly stops matching the standard. Somebody adds a static route during an incident and forgets to remove it. Somebody changes a description string. Somebody disables an unused interface, but only on the switches reachable that night. The configuration has drifted from the intended state.
Drift is dangerous precisely because it is invisible. The network still works after a change, so nothing flags the problem — until you apply the next change and it behaves differently on one device than on the other 199. The remedy is continuous reconciliation: a tool regularly reads each device’s actual configuration, compares it with the intended configuration, and reports or repairs any difference.
Idempotency — the core idea
Traditional CLI configuration is imperative: you issue commands that describe how to change the device, and each command performs an action. If you run the same sequence twice, the second run may fail (“VLAN already exists”) or, worse, do something you did not intend.
Automation tools work differently: you describe the desired end state, and the tool figures out the commands needed to reach it.
| Imperative CLI | Idempotent / declarative | |
|---|---|---|
| You specify | How to change the device (commands) | The state the device must be in |
| Re-running | May error or duplicate work | Safe; makes no change when already correct |
| Drift handling | None — what is there stays there | Detected and corrected on the next run |
| Example | vlan 10 then name DATA | “VLAN 10 named DATA must exist” |
CI/CD at a high level
Software teams solved a very similar problem decades ago: how to make frequent changes safely. The answer was CI/CD — continuous integration and continuous delivery. Network teams have borrowed the same pipeline, treating the network configuration as code.
- Version control — the intended configuration lives in a repository (usually Git). Every change has an author, a timestamp, and a reason.
- Continuous integration (CI) — when a change is proposed, automated checks run: syntax/lint, policy tests, and dry-run comparisons against current device state. The change is validated before it ever touches a device.
- Continuous delivery/deployment (CD) — once validated, the change is applied automatically, often first to a lab or a small pilot group, then to the whole network.
- Verification — after deployment, automated tests confirm the network still behaves as expected. If they fail, the change can be rolled back.
The practical effect is that a network change becomes a reviewable pull request rather than a midnight SSH session. Peers can review it, automated tests can reject it, and if it does go wrong, reverting the commit restores the previous intended state and the next run pushes that state back out to the devices.
The automation tool landscape
You do not need to master any tool for the CCNA, but you should know that tooling falls into three broad families. The rest of this week covers each in turn.
| Family | What it does | Examples later this week |
|---|---|---|
| Controllers | Centralise management and drive devices through APIs. | Cisco DNA Center / Catalyst Center, SDN controllers |
| Configuration-management tools | Declaratively push and enforce device state. | Ansible, Puppet, Chef |
| Scripts and APIs | Talk to devices directly over REST or SSH. | Python, REST APIs, JSON |
The next lesson contrasts the traditional per-device model with a controller-based model, which is the architectural change that makes most modern automation possible. Keep the four benefits (consistency, scale, speed, fewer errors) and the two definitions (drift, idempotency) in mind: everything else in Domain 6.0 is built on them.
Review
Check what you have learned with these review questions.
| Key term | Meaning |
|---|---|
| Network automation | Using software to configure, provision, verify, and operate devices with minimal human interaction. |
| Configuration drift | Unintended divergence of a device’s actual configuration from its intended configuration. |
| Idempotency | Running an operation repeatedly yields the same result as running it once. |
| Declarative | Describing the desired end state; the tool determines the commands to reach it. |
| Imperative | Describing the exact steps or commands to perform; order matters and re-running may fail. |
| CI/CD | Continuous integration (test/validate) and continuous delivery/deployment (apply) of changes. |
| Source of truth | The authoritative intended configuration held in version control. |
| Version control | A system (commonly Git) that tracks every change to configuration code. |
- Cisco. CCNA (200-301) Exam Topics, Domain 6.0 Automation and Programmability. cisco.com.
- Cisco. Cisco DevNet: Network Automation (learning track). developer.cisco.com.
- Hunt, D. & others. Git documentation: About version control. git-scm.com/doc.
- Red Hat. What is CI/CD? redhat.com.
Sources & references
- Cisco. CCNA (200-301) Exam Topics, Domain 6.0 Automation and Programmability. cisco.com.
- Cisco. Cisco DevNet: Network Automation (learning track). developer.cisco.com.
- Hunt, D. & others. Git documentation: About version control. git-scm.com/doc.
- Red Hat. What is CI/CD? redhat.com.
Target exams: CCNA 200-301 v1.1 · CyberOps Associate 200-201. Last content verification: 2026-10-11.