What is serverless computing

In serverless, you write individual functions that respond to events: an HTTP request, a message in a queue, a file uploaded to a bucket or a scheduled cron. The cloud provider handles running the function in an ephemeral container, allocating the necessary resources and destroying the container when it finishes.

The main platforms are AWS Lambda (the pioneer, launched in 2014), Google Cloud Functions, Azure Functions and Cloudflare Workers. Each supports multiple languages (Node.js, Python, Go, Rust, Java) and integrates natively with its provider’s service ecosystem.

The most direct benefit is cost: if your function does not execute, you do not pay. An API receiving 1,000 requests per day can cost less than $1 per month on Lambda. For intermittent or unpredictable workloads, serverless is dramatically more affordable than keeping servers running 24/7.

Eliminating infrastructure management frees the team to focus on business logic. There are no OS security patches, no auto-scaling group configuration, no disk or memory monitoring. The provider manages everything beneath your code.

The cold start is the most well-known limitation. When a function has not been executed recently, the provider needs to create a new container, load the runtime and your code. This cold boot can add between 100ms and several seconds of latency, depending on the language and package size.

Other limitations include maximum execution time (15 minutes on Lambda, 9 minutes on Cloud Functions), deployed package size, available memory and persistent connections (databases, WebSockets). For long-running processes or consistent low-latency requirements, serverless may not be the best option.

Serverless shines for APIs with variable traffic, asynchronous event processing, scheduled tasks (cron jobs) and on-demand data processing. A backend API for a mobile app, an image processing pipeline, or an email notification system are perfect use cases.

It is also excellent for MVPs and prototypes: you can build a functional backend in hours, deploy it without configuring infrastructure and pay only pennies while validating the idea. If the project scales, you can optimise later.

Containers (Docker + Kubernetes) and serverless solve different problems. Containers offer full control over the execution environment, are portable across providers and have no execution time limits. Serverless eliminates infrastructure management but sacrifices control and portability.

For predictable, continuous workloads (a web server with constant traffic), containers tend to be more affordable. For intermittent workloads with spikes (an API receiving 100 requests per hour and 10,000 during a campaign), serverless scales better and costs less. Many modern architectures combine both: containers for core services and serverless for auxiliary functions.

The serverless ecosystem has matured with frameworks that simplify development and deployment. Serverless Framework, SST (Serverless Stack) and AWS SAM are the most popular. For edge computing, Cloudflare Workers and Vercel Edge Functions run code at points of presence close to the user with microsecond cold starts.

For local development, tools like LocalStack emulate AWS services, and each framework has its own local development mode. Observability is critical: Datadog, Lumigo and AWS X-Ray enable tracing invocations and detecting bottlenecks in serverless architectures.