Case Study: FOSS (Apache Hbase) at Flipkart
summary
Flipkart describes how HBase is customized to run 5 million QPS on Kubernetes by keeping core code untouched while adding multi-tenant features: per-tenant region-group multi-tenancy, storage-class isolation (NVMe for hot 1TB tenants, HDD for cold 100+TB tenants) and WAL isolation so each tenant writes its own write-ahead-log. Balancer plugins and cost functions enforce tenant-aware region distribution; upgrades follow HBase 1.x → 2.1x → 2.5x with patchable, testable changes.
AI-written summary · about
chapters (10)
- 0:00FOSS at Flipkart
- 1:53Why Flipkart relies on open source
- 3:42Open source advantages
- 7:18Flipkart’s stack and HBase
- 8:43Choosing HBase
- 11:50Customizing HBase for multi-tenancy
- 14:44Data layer isolation in HBase
- 17:58Scaling HBase dynamically
- 19:20Long-term benefits of open source
- 20:04Q&A
related talks
29mMulti Tenancy — Apache Hbase
49mHow Flipkart runs HBase across zones and clouds seamlessly
50mRunning Databases Efficiently at Scale - Case Study of Flipka...
29mFlipkart: Scaling Aerospike at Flipkart with Kubernetes
43mMulti-tenant HBase Solutions at Facebook
11mFOSS @ ElasticRun
55mApache HBase: crazy dances on the elephant back
32mContaining an Elephant