Build Durable Applications with Temporal

An hour on why modern applications fail in boring, preventable ways, and what changes when your business logic can survive the process it's running in.

Introduction

We broke the monolith into services because the monolith was hard to change. What we got back was a network between every function call, and a whole category of failure that didn't exist before: the partial success, the timeout you can't distinguish from a slow response, the retry that charged the customer twice.

Most teams answer this with the same scaffolding every time. A queue, a state table, a reconciliation job, a retry loop with backoff someone wrote at 2am. That code is not the product, but it's most of the codebase.

Temporal makes durability a property of the runtime instead of something you hand-roll. The talk walks through a small application, runs it, kills it mid-execution, and watches it pick up exactly where it left off.

Delivery

This talk was given at:

• Austin Python User Group, March 13, 2023
• Temporal Austin Meetup at SailPoint, July 22, 2024

Adapted from Brian Hogan's presentation, with the examples ported to Python.