Field Guide · concept

Also known as: FaaS, Functions as a service

Serverless computing is a cloud computing model in which code runs in short-lived, provider-managed environments triggered by events; it scales automatically and bills only for actual execution, with no servers to provision.1

HTTP request queue msg timer events functions spun up on demand fnfnfnfn idle scale to zero you pay only while a function is actually running
Each event fires a function that the platform runs in a freshly created container; a burst spins up many in parallel, and when events stop the platform scales to zero — so idle time costs nothing.

Overview

The name is a slight misnomer — servers still exist, but the developer never sees or manages them. You upload a function; the platform runs it on demand when an event fires (an HTTP request, a queue message, a timer), spins up containers to handle load, and tears them down afterward. Because idle functions cost nothing, serverless can scale to zero between bursts.

The trade-offs are real: a cold start delay when a new container must be created for the first request, hard limits on how long a single invocation may run, and statelessness — each call starts fresh, so anything that must persist lives in an external store. Long-running or stateful work fits poorly, but bursty, event-driven tasks fit beautifully.

Trade-offs

Serverless sits at the top of the cloud abstraction ladder, giving up control for convenience:

Property Serverless Traditional server
Provisioning None You size the machine
Scaling Automatic, to zero Manual or scripted
Billing Per execution Per hour, even idle
Startup latency Cold-start delay Always warm
Long / stateful jobs Poor fit Fine

Where the workload is spiky and stateless, paying only for execution is a big win; where it is continuous, a server that stays warm is both simpler and cheaper.

Where it fits

Serverless is the most abstract way to run your own code, sitting above platform as a service by removing even the notion of a running process you manage. Its automatic scalability suits bursty, event-driven workloads. GopherTrunk’s decode loop is continuous and tied to live RF, so it is a poor fit for serverless, but occasional tasks — sending an alert when a talkgroup appears — map naturally onto a function.

Sources

  1. Serverless computing — Wikipedia, on the serverless and FaaS model. ↩

See also