Dream11: Aerospike Beyond Cache
summary
Dream11 replaces Amazon RDS and Elasticsearch with Aerospike as a single primary store, adds TTL-based auto-archival, and implements smart pagination using a month-level shard meta get plus mget to avoid scans. The system endures extensive chaos testing (node failures, split brain, EBS IO faults) and supports 5 million writes per second and ~1.3 million reads per second for team service; IPL load shows ~0.5M reads and 50k writes with P99 ≈1 ms while using linearized reads for strong consistency.
AI-written summary · about
chapters (16)
- 0:00Dream11 and Aerospike intro
- 1:00Aerospike beyond cache
- 4:09Old architecture problems
- 7:12New system design goals
- 8:14New Aerospike architecture
- 9:27Smart pagination with Aerospike
- 11:30IPL 2023 final metrics
- 12:42Team service and chaos testing
- 18:33Q&A
- 20:12Hotkey issues in Aerospike
- 21:46Old vs new architecture
- 22:53Linearized reads use case
- 23:43RDS to Aerospike migration
- 25:17XDR to Kafka for Aerospike
- 26:51Cluster switch conflict handling
- 27:34Team service upgrade and uptime
related talks
25mCash the Cache - How BigBasket uses Aerospike
48mThe real-time database to build your AI future
49mJourney with Aerospike: Architecture Best Practices
15mAWS On Air Voice of the Customer ft. Aerospike, Inc.
31mAerospike's Innovation Agenda
22mThe Art of the Possible
33mInnovation with Aerospike: How AppLovin Handles Billions of Requests per Day
31mHow Dream11’s Fantasy Platform Is Handling The 2021 Indian Premier League