FlightTracker: Social Graph Consistency at Scale
summary
FlightTracker enforces user-centric read‑your‑writes by issuing compact tickets that list recent object keys and versions and attaching them to reads so stores perform ticket-inclusive reads. Global secondary indexes use a complementary metadata service called Arrivals to know whether index updates have applied; if stale, queries are re-evaluated on fresh replicas. Production numbers: >100M read QPS and >20M write QPS through FlightTracker, >1 quadrillion store/index queries/day, <2% CPU/RAM overhead, six nines read availability and 10× write availability improvement.
AI-written summary · about
chapters (9)
- 0:00Social graph consistency
- 1:06Graph challenges
- 6:50FlightTracker overview
- 9:14Ticket mechanism
- 12:21Ticket inclusive reads
- 14:27Global secondary indexes
- 17:40Production impact
- 18:43Extending guarantees
- 21:37Deployment lessons
related talks
52mTaking Storage for a Ride
44mScanning the skies above SQLBits - Realtime streaming of flight data with Microsoft Fabric
30mTransUnion: Scaling Data Heavy Transactions w/ Aerospike + Adaptive Map
54mBreakout Session
34mDistributed Tracing, Lessons Learned
50mAerospike Technical Overview
41mInside Twitter's Realtime Delivery of Tweets for Enterprise Customers
28mUsing a database in a data science project - Lessons learned in production