Fail Fast - Tumblr's Evolving Architecture

Explore Tumblr's evolving architecture in this 52-minute conference talk from Philly ETE 2013. Dive into the challenges faced by the popular blogging platform as it scaled to handle massive traffic growth. Learn about the various architectural changes implemented from 2010 to 2013, including strategies for managing aggregated data, improving blog and dashboard views, and handling large-scale batch processing. Discover the tools and processes Tumblr adopted, such as Jetpants for automated MySQL topology management, developer-driven deploys, and Collins for automated provisioning. Gain insights into monitoring techniques, open-sourced solutions, and valuable lessons learned from Tumblr's journey to build a scalable and resilient infrastructure.

Intro
What's Tumblr?
Stats Today
Traffic Growth
App-Based Challenges
Procedural Challenges
Plateaus of Scalability
2010 Architecture
Additional Aggregated Data
2011 Architecture
2012 Architecture
Blog View
The Ceiling
Dashboard View
2013 (In-progress) Architecture
Data Volume Assumptions
Large Scale Batch Processing
Team Growth
Jetpants - Automated MySQL Topology Management
Dev Cycle
Developer Driven Deploys
Ramped Deploys
Trending
Monitoring
Collins: Automated Provisioning
Everything Ties to Collins
Open Sourced!
Open Challenges
Scalable Tools Tools We Like
Lessons Learned