Automation & Programmability
CCNA 200-301 · Automation and Programmability
Automation is the newest domain on the CCNA and the one that most surprises candidates who expected a purely command-line exam. It is only 10% of the test, and it is almost entirely conceptual — no programming is required, and no code needs to be written. The exam wants you to understand why networks are being automated, how controller-based networking differs from traditional networking, what a REST API is, how to read JSON data, and what tools like Ansible and Terraform do. This guide covers those ideas in plain language, with the vocabulary the exam uses.
How automation impacts network management
Traditional network management means logging into devices one at a time and typing commands. It works for ten devices; it does not work for ten thousand. Every manual change is slow, inconsistent (two engineers configure the same task slightly differently), and error-prone (a typo can take down a network). Automation replaces repetitive manual work with software: a script or a tool applies the same tested configuration to every device, in seconds, with the same result every time. The benefits the exam names are consistency, speed, reliability, and scale — plus the ability to treat network configuration as code that can be reviewed, version-controlled, and rolled back, exactly like application code. Automation does not replace network engineers; it replaces the repetitive parts of their jobs and frees them for design and troubleshooting.
Traditional networks vs controller-based networking
In a traditional network, every device makes its own decisions. Each router runs its own routing protocol, builds its own routing table, and forwards packets based on its own view of the network. The control plane (the decision-making: which path should traffic take?) and the data plane (the forwarding: actually moving the packets) live together on every device, and changes are made device by device. In controller-based networking (the core idea of software-defined networking, or SDN), the control plane moves to a central controller, and the network devices become simpler: they forward packets according to instructions from the controller. The controller sees the whole network, so it can make decisions no single device could — rerouting traffic around a failure in seconds, or applying a security policy to every device at once.
SDN architecture has three parts the exam names. The underlay is the physical network: the routers, switches, and links that carry traffic. The overlay is the virtual network built on top of the underlay: tunnels and virtual segments that connect endpoints regardless of the physical topology. The fabric is the integrated combination of underlay and overlay, managed as one system — Cisco's SD-Access is a fabric architecture for campus networks. Two types of API connect the pieces: southbound APIs carry instructions from the controller down to the network devices, and northbound APIs let applications and orchestration systems talk to the controller, asking it for information or requesting network changes. Remember the direction: north is up toward the applications, south is down toward the devices.
AI and machine learning in network operations
The v1.1 blueprint added AI to the exam, and v2.0 expands it further — so this topic is worth knowing properly. Predictive AI and machine learning analyse network data to forecast problems before they happen: spotting a traffic pattern that precedes a failure, predicting when a link will saturate, or flagging behaviour that looks like an attack. Generative AI creates new content: drafting a configuration from a plain-language request, summarising a troubleshooting session, or answering an engineer's question in natural language. In practice the two combine in digital network assistants and AIOps platforms, which monitor the network continuously, explain what they find, and sometimes suggest or apply fixes. The exam's point is the distinction — predictive AI forecasts, generative AI creates — and the idea that AI augments engineers rather than replacing them: a human still approves changes and is accountable for the result.
REST-based APIs
An API (Application Programming Interface) is how software talks to other software, and a REST-based API is the dominant style for network automation. REST uses standard HTTP requests to perform operations on resources, and the exam maps the four CRUD operations to HTTP verbs: Create uses POST, Read uses GET, Update uses PUT or PATCH, and Delete uses DELETE. A GET request to a controller's API might retrieve the list of devices; a POST might create a new VLAN. REST APIs are stateless — each request carries everything the server needs, and the server remembers nothing between requests — and they exchange data in a structured format, usually JSON.
APIs need authentication, and the exam names three types: basic authentication (a username and password sent with the request, only safe over HTTPS), API keys (a long token identifying the caller, sent in a header), and bearer tokens (a token obtained by logging in first, then presented with each request — the OAuth style). The API's responses carry HTTP status codes that tell you what happened: 200 means OK, 201 means a resource was created, 400 means the request was malformed, 401 means authentication failed, and 404 means the resource was not found.
Configuration management: Ansible and Terraform
Ansible and Terraform are the two automation tools the exam names, and they do different jobs. Ansible is a configuration management tool: it connects to devices over SSH (no agent software needed on the device — it is agentless), and it applies configurations described in YAML files called playbooks. An Ansible playbook might say "make sure every switch has this NTP server configured", and Ansible checks each device and changes only what is needed. Terraform is an infrastructure-as-code tool: it provisions and manages infrastructure — virtual machines, networks, cloud resources — from declarative configuration files written in HCL, tracking what exists in a state file and applying only the changes needed to reach the desired state. The exam's distinction is simple: Ansible configures things that already exist, Terraform creates things. Both let a team manage infrastructure with version-controlled files instead of manual steps.
JSON-encoded data
JSON (JavaScript Object Notation) is the data format REST APIs use, and the exam asks you to recognise its components. JSON data is built from objects (key-value pairs inside curly braces), arrays (ordered lists inside square brackets), strings (text in double quotes), numbers, booleans (true/false), and null. A typical example: {"hostname": "R1", "interfaces": ["GigabitEthernet0/0", "GigabitEthernet0/1"], "ospf_enabled": true}. The keys are strings, the values can be any type, and the structure nests — an object can contain arrays, and arrays can contain objects. You do not need to write JSON on the exam, but you do need to read it: given a JSON snippet, identify the keys, the values, and the data types. The most common mistakes are using single quotes instead of double quotes (JSON requires double quotes) and forgetting that a trailing comma after the last item is not allowed.
Exam traps to watch for
- Saying the controller forwards user traffic. The controller makes decisions (control plane); the devices forward packets (data plane).
- Mixing up northbound and southbound APIs: northbound connects the controller to applications above it; southbound connects the controller to devices below it.
- Confusing Ansible and Terraform: Ansible configures existing devices (agentless, SSH, YAML playbooks); Terraform provisions infrastructure (declarative, state files).
- Mapping CRUD verbs wrongly: Create is POST, Read is GET, Update is PUT/PATCH, Delete is DELETE.
Practice questions
Test yourself on this guide. Answers are checked instantly, with a full explanation for each question.
Q1.What is the main benefit of network automation over manual device-by-device configuration?
Explanation: Automation applies the same tested configuration to every device quickly and identically, which removes the inconsistencies and typos of manual configuration and scales to networks far too large to manage by hand. Option A is wrong because automation changes the engineer's job — removing repetitive work — rather than removing the engineer. Option C is wrong because automation does not change physical link speeds. Option D is wrong because routing protocols still run on the devices; automation changes how they are configured, not whether they are needed.
Exam objective: 200-301 6.1: Explain how automation impacts network management
Q2.In a controller-based (SDN) architecture, what is the difference between the control plane and the data plane?
Explanation: The control plane is the decision-making intelligence (which path should traffic take?), centralised in the controller in an SDN design; the data plane is the actual forwarding of packets, which happens on the network devices following the controller's instructions. Option B reverses the two. Option C is wrong because both planes are logical functions, not physical media. Option D is wrong because the separation of control and data planes is the defining idea of SDN.
Exam objective: 200-301 6.3: Describe controller-based, software defined architecture — separation of control plane and data plane
Q3.In an SDN architecture, which API lets an application request a network change from the controller, and which API carries the controller's instructions to the devices?
Explanation: Northbound APIs face upward from the controller toward applications and orchestration systems, while southbound APIs face downward from the controller toward the network devices it manages. Option A reverses the directions. Option B is wrong because the two directions use different APIs with different purposes. Option D is wrong because southbound APIs go to devices, not applications, and northbound APIs are not controller-to-controller links.
Exam objective: 200-301 6.3: Describe controller-based, software defined architecture — northbound and southbound APIs
Q4.A network operations tool analyses traffic patterns and warns that a WAN link will saturate next Tuesday. Is this generative AI or predictive AI, and why?
Explanation: Predictive AI and machine learning analyse historical data to forecast future events — here, predicting link saturation before it happens. Option A is wrong because generating a text warning is not what defines generative AI; the forecasting is the key function. Option C is wrong because the tool is not creating a traffic pattern. Option D is wrong because writing a configuration would be a generative or automation action, not a prediction.
Exam objective: 200-301 6.4: Explain AI (generative and predictive) and machine learning in network operations
Q5.Which HTTP verb is used to retrieve information from a REST API, and which is used to create a new resource?
Explanation: In the CRUD mapping for REST APIs, Read uses GET and Create uses POST (Update uses PUT or PATCH, Delete uses DELETE). Option B swaps the two verbs. Option C is wrong because PUT updates and DELETE removes; neither retrieves or creates. Option D is wrong for the same reason.
Exam objective: 200-301 6.5: Describe characteristics of REST-based APIs (authentication types, CRUD, HTTP verbs, and data encoding)
Q6.A REST API returns status code 401 in response to a request. What does this mean?
Explanation: HTTP status code 401 Unauthorized means the request failed authentication — the credentials or token were missing or invalid. Option A is wrong because a successful request returns 200 OK. Option B is wrong because resource creation returns 201 Created. Option C is wrong because a missing resource returns 404 Not Found.
Exam objective: 200-301 6.5: Describe characteristics of REST-based APIs (authentication types, CRUD, HTTP verbs, and data encoding)
Q7.An engineer wants to configure NTP on 200 existing switches without installing any software on them. Which tool fits best, and how does it connect?
Explanation: Ansible is agentless — it connects to existing devices over SSH with no software to install on them — and it applies configurations described in YAML playbooks, which fits the task of configuring NTP on 200 switches. Option A is wrong because Terraform provisions new infrastructure rather than configuring existing devices. Option C is wrong because a REST API is an interface, not a configuration management tool. Option D is wrong because JSON is a data format, not a tool.
Exam objective: 200-301 6.6: Recognize the capabilities of configuration management mechanisms such as Ansible and Terraform
Q8.In the JSON data {"hostname": "R1", "interfaces": ["G0/0", "G0/1"], "ospf_enabled": true}, what is the data type of the value for "interfaces"?
Explanation: The value ["G0/0", "G0/1"] is enclosed in square brackets, which makes it an array — an ordered list of values, here two strings. Option A is wrong because a string is a single piece of text in double quotes, like the value of "hostname". Option B is wrong because a boolean is true or false, like the value of "ospf_enabled". Option C is wrong because there is no number in this example.
Exam objective: 200-301 6.7: Recognize components of JSON-encoded data
Q9.In a software-defined architecture, what is the difference between the underlay and the overlay?
Explanation: The underlay is the physical infrastructure — routers, switches, and links — while the overlay is the virtual network of tunnels and segments built on top of the underlay, connecting endpoints regardless of the physical topology. Option B is wrong because the controller and firmware are software components, not the underlay/overlay distinction. Option C is wrong because the distinction has nothing to do with wired versus wireless. Option D is wrong because northbound and southbound describe API directions, not network layers.
Exam objective: 200-301 6.3: Describe controller-based, software defined architecture (overlay, underlay, and fabric)
Q10.How does controller-based networking differ from traditional networking?
Explanation: In traditional networking, every device runs its own control plane — its own routing protocol and forwarding decisions. In controller-based (SDN) networking, the control plane moves to a central controller that sees the whole network and instructs the devices, which focus on forwarding (the data plane). Option A describes traditional networking, not controller-based networking. Option C is wrong because devices still need IP addresses in both architectures. Option D is wrong because the centralisation of the control plane is the defining difference between the two models.
Exam objective: 200-301 6.2: Compare traditional networks with controller-based networking