Wouter de Bie - BigDataNYC 2013 - theCUBE - #BigDataNYC
summary
Spotify builds a company-wide data platform, early adopting Hadoop + Hortonworks to handle millions of users and more than 20 million tracks, running on on‑prem clusters for low‑latency playback. Analysts use Hive for ad‑hoc queries; the stack is migrating to HDP2/YARN so teams can run graph processing and non‑MapReduce ML instead of forcing problems into map‑reduce. Recommendations run with collaborative filtering; A/B testing and product instrumentation drive feature decisions and faster iteration.
AI-written summary · about
