How Flipkart runs HBase across zones and clouds seamlessly
summary
Flipkart explains how HBase clusters span regions and hybrid clouds by using HDFS and HBase WALs for ordered replication, relying on async replication and cluster UUIDs to prevent cycles, and layering a client wrapper open sourced that implements replica sets, routing policies and intent store for payments plus CDC that pushes filtered changes to Kafka/Pulsar. Deployment mixes active‑active and active‑passive topologies; more than 200+ use cases run on this stack, and replication plus client-side routing provides eventual consistency with recoverable intent and CDC pipelines.
AI-written summary · about
chapters (10)
- 0:00Introduction to HBase
- 1:20HBase architecture and layout
- 3:13HBase write and replication
- 5:28Multi-DC deployment strategies
- 11:25Replication and conflict handling
- 20:19Client-side basics
- 28:14Client routing mechanisms
- 37:05Handling intent writes
- 41:32Change data capture (CDC)
- 44:53Conclusion and takeaways
related talks
21mCase Study: FOSS (Apache Hbase) at Flipkart
29mMulti Tenancy — Apache Hbase
29mFlipkart: Scaling Aerospike at Flipkart with Kubernetes
50mRunning Databases Efficiently at Scale - Case Study of Flipka...
06mLearning curve with HBase deployment
55mApache HBase: crazy dances on the elephant back
32mContaining an Elephant
41mBuilding tiered data stores using Aesop to bridge SQL and NoSQL systems