Merging modern high-scale AdTech/MarTech at the edge
summary
AcuityAds uses Aerospike at the edge with CDTs and Lua UDFs to implement a real-time journey state machine that enforces frequency caps, user matching, consent and bid token caching; the cluster holds 60 billion records, sustains 3.7M reads/sec and 150k writes/sec with ~1 ms DB latency and ~100 ms end‑to‑end SLA. Tokens use protobuf, short‑TTL memory namespaces promote hot/cold movement to SSD, replication factor 2 and XDR reduce split‑brain to under 1%, and the Lua UDFs provide the needed atomicity.
AI-written summary · about
chapters (28)
- 0:00Introduction
- 1:03About AcuityAds
- 2:50Technology overview
- 3:51Hbase
- 4:33Latency
- 5:28System Context
- 6:36Cluster Overview
- 7:45Cookie Syncing
- 9:16Privacy
- 10:00Exchange IDs
- 10:21User Segmentation
- 11:11User Segmentation Model
- 12:24Frequency Capping
- 13:19Record Structure
- 13:55Client API
- 14:37Bid Token Caching
- 16:53Illumina
- 17:31UserDefined Canvas
- 19:04Journey State Machine
- 20:24Journey Topology
- 21:30Initial Step Cache
- 22:35Journey Transition Service
- 23:25Design Decisions
- 24:23Journey Record
- 25:11User Journey Record
- 25:51Audience Engagement Map
- 26:55Conclusion
- 27:24Outro
related talks
01h 02mAchieving High Load in Advertising Technology
40mMastering small chunks of data in high frequency ad-tech systems
26mPowering Real-time Decisions With a Reduced Footprint
03mInterview
39mEdge Compute Infrastructure for Experience Business
33mUsing Aerospike as a Real-time System-of-Record for Adobe Experience Cloud Services on Azure
23mAdobe + Aerospike: Real-time Streaming Segmentation
08mInMobi: Powering Supply Path Optimization with Aerospike