Multi-Sided Marketplace & Data PipelinesPRODUCTION RETROSPECTIVE

Airbnb: Strangler Fig Monorail Migration & Zipline ML Feature Store

Scaling marketplace search and dynamic pricing: Monorail migration using Debezium CDC and Strangler Fig, calendar lock concurrency, and Zipline offline-to-online ML feature stores.

High-Level Architectural Overview

Airbnb successfully decomposed its legacy monolithic Ruby on Rails application ("Monorail") into hundreds of microservices using the Strangler Fig pattern paired with Change Data Capture (CDC) over Kafka, ensuring zero downtime and zero data inconsistency during the multi-year migration.

Key Engineering Problems & Trade-Offs

Strangler Fig Monorail Migration with CDC

Millions of global bookings, hundreds of microservices
The Scaling Problem

Decomposing a massive legacy monolithic Rails codebase without stopping feature development or risking transactional data inconsistencies.

Engineering Solution

Strangler Fig pattern: new microservices are placed behind an API proxy. Dual-writes and Debezium Kafka Change Data Capture (CDC) ensure bidirectional data sync until the monolith is safely deprecated.

Architectural Trade-Offs

Temporary dual-write complexity and shadow traffic validation overhead vs zero downtime migration.

How to Say This in an Interview

When discussing monolith-to-microservices migration in interviews, emphasize the Strangler Fig pattern and Change Data Capture rather than big-bang cutovers.

Curriculum Topics Used in Airbnb Architecture (5)
Full Syllabus
Primary Technical Sources & Published Papers

Ready to Practice Airbnb-Style Systems?

Start with foundational networking, compute, and storage, and build up to complex distributed consensus.

Start Free: Topic #1