What is FaaS (Function-as-a-Service)?
Learn what FaaS is, how event-triggered functions and cold starts work, and how it differs from broader serverless models.
Expected Interview Answer
FaaS (Function-as-a-Service) is a serverless execution model where individual, single-purpose functions are deployed independently and invoked automatically in response to events, with the provider handling all provisioning, scaling, and isolation between invocations.
Each FaaS function is stateless and short-lived by design: it starts, processes one event such as an HTTP request, a message queue entry, or a database change stream, returns a result, and the execution environment is then either frozen for reuse or torn down entirely. Because functions are billed per invocation and per millisecond of execution, FaaS is the finest-grained unit of serverless compute, in contrast to broader serverless platforms that might run entire containers or services on demand. State that needs to persist across invocations must be stored externally, in a database, object store, or cache, since the local execution environment cannot be relied upon to survive between calls. FaaS platforms like AWS Lambda, Azure Functions, and Google Cloud Functions typically enforce a maximum execution duration and impose cold-start latency on infrequently invoked functions, which is why FaaS suits short, event-driven tasks far better than long-running processes.
- Deploys individual functions independently with no server management
- Bills at fine-grained per-invocation, per-millisecond granularity
- Scales automatically per function from zero to thousands of concurrent calls
- Simplifies event-driven architectures like webhooks and stream processing
AI Mentor Explanation
FaaS is like a stadium hiring a single specialist, such as a boundary-rope official, who is called onto the field only for the exact moment a ball nears the boundary, does that one job, and then steps back off. Each time a ball approaches the rope is a separate event, and a fresh official assessment happens independently every single time with no memory of the previous boundary call carried over. If any decision needs to be remembered for later, like a running tally, it has to be written down on the scoreboard rather than kept in the official’s head. This lets the stadium staff many tiny specialist roles without keeping every official on the field for the entire match.
Step-by-Step Explanation
Step 1
Deploy a single-purpose function
Package one narrow function independently, without a broader server or container to manage.
Step 2
Register the trigger
Bind the function to an event source: HTTP request, queue message, or stream/database change.
Step 3
Execute on invocation
The provider allocates an isolated environment, runs the function once, and returns the result.
Step 4
Persist state externally
Any state needed beyond that invocation is written to a database, cache, or object store, never kept locally.
What Interviewer Expects
- Understanding that FaaS functions are stateless and short-lived by design
- Knowledge that FaaS is the finest-grained unit of serverless compute
- Awareness of billing per invocation/millisecond versus provisioned servers
- Ability to name real platforms (Lambda, Azure Functions, Cloud Functions) and their constraints
Common Mistakes
- Treating FaaS as identical to all of “serverless” rather than one specific model within it
- Assuming state persists in memory across separate invocations
- Ignoring maximum execution duration limits when designing long tasks
- Not distinguishing FaaS functions from serverless container platforms
Best Answer (HR Friendly)
“FaaS lets us write small, single-purpose functions that run automatically whenever a specific event happens, like an API call or a new file upload, and we only get billed for the exact milliseconds that function runs. It is a great fit for lightweight, event-driven tasks where we do not want to manage or pay for a server sitting idle the rest of the time.”
Code Example
# Deploy a function bound to an HTTP trigger
gcloud functions deploy processUpload \
--runtime=nodejs20 \
--trigger-http \
--entry-point=handler \
--memory=256MB \
--timeout=30s
# Invoke it directly to test
curl -X POST https://REGION-PROJECT.cloudfunctions.net/processUpload \
-H "Content-Type: application/json" \
-d '{"fileId": "abc123"}'Follow-up Questions
- How does FaaS differ from running a container on a serverless platform?
- Where should state be stored if it needs to persist between invocations?
- What happens if a FaaS function exceeds its maximum execution duration?
- How would you design a FaaS-based pipeline for processing uploaded files?
MCQ Practice
1. What is a core characteristic of a FaaS function?
FaaS functions are designed to be stateless and short-lived, invoked once per event and then torn down or frozen.
2. Where should data be stored if it must persist across separate FaaS invocations?
Since the execution environment is not guaranteed to survive between invocations, persistent state must live in an external store.
3. How is FaaS typically billed?
FaaS platforms bill at fine granularity based on the number of invocations and the execution duration/resources consumed.
Flash Cards
What is FaaS? — A serverless model running single-purpose, event-triggered functions billed per invocation.
Are FaaS functions stateful? — No — they are stateless and short-lived; persistent state must live externally.
Name two real FaaS platforms. — AWS Lambda and Azure Functions (also Google Cloud Functions).
What limits FaaS for long tasks? — A maximum execution duration enforced by the platform, plus cold-start latency.