2learn

Serverless Computing

What it is

Serverless computing is running code on a managed runtime that starts it when a request arrives and stops it when the request is done. There is no machine to patch and nothing to pay for while the code is idle; the provider sizes, starts and retires the runtime. The name is misleading — there are servers, they are simply not your problem — and the useful mental model is “a function plus the managed services it calls”, because a function by itself does very little.

Why it matters

For spiky and intermittent work, serverless is dramatically cheaper: a task triggered a few times a day costs nothing the rest of the time, where a server would run all month. It also removes an entire class of operational work, which matters most to small teams without a platform group. The trade is that you inherit the platform’s constraints — timeouts, memory ceilings, cold starts, no local disk worth speaking of — and that debugging is harder when you cannot log into the machine. It is a good fit for glue, event handling and bursty APIs, and a poor fit for long-running computation or anything needing unusual system access.

In this track

  1. Functions and managed runtimes (06-cloud-devops/05-serverless-computing/01-functions-and-managed-runtimes) — what a function is, and the services it is actually assembled with.
  2. Designing for serverless (06-cloud-devops/05-serverless-computing/02-designing-for-serverless) — the constraints that decide whether it fits, and the patterns that work within them.

Prerequisites

Go deeper

Analytics cookies are set only if you allow them. Nothing is loaded, and nothing is sent, until you choose.