Mastering small chunks of data in high frequency ad-tech systems
summary
Virtual Minds uses Aerospike as the primary low‑latency key‑value store for tagging, DMP segments and a DSP bidder, handling peaks to 1M requests/sec across ~300M profiles. Enterprise features chosen are unlimited nodes plus fast restart and cross‑DC replication. Typical SLA is 95% <5ms and external‑ID lookups average ~1ms; budget control survives million‑QPS spikes. Deployment is on‑prem with hybrid/kubernetes plans, Kafka/Spark connectors in POC, and operations remain lightweight.
AI-written summary · about
chapters (19)
- 0:00Introduction
- 3:25Virtual Minds overview
- 7:46Lessons learned
- 12:04What led to Aerospike
- 15:03Patterns
- 20:33Bidding example
- 21:43Budget control
- 23:04Traffic
- 24:42Identity Mappings
- 25:39Aerospike Enterprise
- 26:28Cross data center replication
- 26:59Make or buy
- 28:01Onpremise
- 29:57Other use cases
- 30:44What makes the enterprise software so valuable
- 31:25What else are you aiming for
- 32:26Whats your view on Kafka Spark Connector
- 32:50Summary
- 34:21Questions
related talks
08mBuilding a Real-time Data Future
28mMerging modern high-scale AdTech/MarTech at the edge
23mAdobe + Aerospike: Real-time Streaming Segmentation
33mUsing Aerospike as a Real-time System-of-Record for Adobe Experience Cloud Services on Azure
23mMyntra's Sub-Millisecond Journey with Aerospike
29mWhat breaks when Aerospike hits 6 million QPS
48mThe real-time database to build your AI future
15mAWS On Air Voice of the Customer ft. Aerospike, Inc.