HBase and Its Associated Services - Data@Scale

Explore Pinterest's journey in architecting and scaling Feed storage using HBase in this 27-minute talk by software engineer Varun Sharma. Gain insights into the critical role of Feeds in user experience, the rationale behind choosing HBase, and the implementation of personalized and following feeds. Delve into various challenges faced, including rich pins, recommendations, and single points of failure. Learn about performance optimization techniques, simulation strategies, and future pipeline plans. Discover how Pinterest tackled Mean Time To Recovery (MTTR) issues and designed a robust Feeds architecture to support one of its most demanding applications.

Intro
Outline
Storage stack @Pinterest
Why HBase ?
Where?
Personalized Feeds
Following Feed on HBase
"Misc" Challenges
Feeds Architecture
Rich Pins - 11
Recommendations
HOW?
MTTR - 11
MTTR - IV
Simulate, Simulate, Simulate
Performance
In the Pipeline...
Single Points of Failure - 11