Aerospike Technical Overview
summary
Claims strong primary-key consistency at high throughput using a roster to define namespace members, a master-oriented hybrid clock, and a bespoke two‑phase commit with an advisory replicated packet that speeds re-sync ~100x. Replication factor 3, linearized reads cost performance but still reach ~1M TPS on modest clusters; certain network splits intentionally return unavailable for safety. Also ships a Spark-integrated real‑time analysis add‑on called Raph (1.4 TB test on AWS using ~32 nodes plus 4 Raph nodes).
AI-written summary · about
chapters (15)
- 0:00AEROSPIKE USER SUMMIT 2018
- 1:06Classic Distributed System Failures
- 6:26Hybrid Clock
- 8:07Aerospike 4.0: Master/Replica Promotion and Availability Roster - This defines the list of nodes that are part of the cluster in steady state
- 11:04Aerospike 4.0: Write Logic (2PC)
- 13:39Aerospike 4.0: Linearizing Reads
- 27:03Aerospike related ecosystems in an Enterprise
- 31:46Real-time Business Insights - Challenges
- 33:05Introducing a new approach for Real-time Insights: Aerospike Realtime Analysis Framework (RAF)
- 35:45Features and Benefits of RAF 1.0
- 37:40A Sample Deployment Scenario
- 38:56Risk Management in Capital Markets
- 40:06Fraud Detection
- 40:44Credit Card Processing Network
- 41:31Personalization & Recommendation Engines
related talks
24mWhat a Modern Database Enables
22mThe Art of the Possible
30mLatest Innovations
49mJourney with Aerospike: Architecture Best Practices
31mAerospike's Innovation Agenda
09mBrian Bulkowski (Aerospike) Interview - OSCON 2014
29mWhat breaks when Aerospike hits 6 million QPS
31mAerospike Today and Tomorrow: Our Roadmap