The InfoQ Podcast

Building a Reliable Kafka Data Processing Pipeline With Lily Mara

10 snips
Aug 14, 2023
Ask episode
AI Snips
Chapters
Transcript
Episode notes
ANECDOTE

OneSignal's Initial Architecture

  • OneSignal initially used a Rails monolith for all operations, including HTTP handling and asynchronous job processing with Sidekick.
  • This setup worked well for sending large message campaigns but encountered scaling challenges with high-throughput HTTP endpoints.
ADVICE

Introduce Queuing for High Throughput

  • Implement queuing to control concurrency and load on Postgres when dealing with high-throughput HTTP endpoints.
  • OneSignal achieved this by switching from synchronous processing in HTTP handlers to a Kafka producer and a Rust consumer.
ANECDOTE

Early Adoption of Rust

  • OneSignal adopted Rust early on for core components like OnePush, valuing its speed, low overhead, and ease of development.
  • They were early adopters, using Rust shortly after its 1.0 release, and have been pleased with its performance and maintainability.
Get the Snipd Podcast app to discover more snips from this episode
Get the app