Real Time Marketing with Kafka, Storm, Cassandra & Spark #bbuzz
summary
A realtime CRM pipeline sends game events through Kafka into Storm where JavaScript handlers update per-day player profiles in Cassandra and emit targeted interstitials; Kafka runs on five nodes with 16 partitions per raw topic and four days retention, the system ingests about 500 million events/day, Storm applies Nashorn JavaScript snippets for runtime-changable logic, Cassandra is an eight-node profile store and Spark with a Cassandra connector answers cross-player analytics; deployed to test markets now, it prevents churn by offering contextually relevant offers and shows promising uplift though full production numbers are pending.
AI-written summary · about
chapters (16)
- 0:00Introduction and speaker intro
- 1:30Real-time CRM motivation
- 3:39Data infrastructure overview
- 7:52Realtime CRM process steps
- 8:56Kafka setup at InnoGames
- 11:25Storm basics and topology
- 13:18Realtime CRM Storm topology
- 15:46JavaScript handlers in Storm
- 17:12Storm alternatives and updates
- 19:41Cassandra data model explained
- 22:54Player profile in Cassandra
- 24:18Spark for analytical queries
- 26:45Spark job server use case
- 27:06Project takeaways and lessons
- 29:43Next steps and future plans
- 30:26Q&A session
related talks
36mReal Time Big Data Analytics with Kafka, Storm, HBase #bbuzz
01h 00mPutting Apache Kafka to Use
59mStreaming Big Data w/Spark, Spark Streaming, Kafka, Cassandra & Akka
38mReal-time analytics with Apache Cassandra & Apache Spark
42m"Distributed stream processing in Ruby"
30mArchitecting Applications With Multiple Open-Source Big Data Technologies
28mKeynote Panel Discussion: SMACK Today and Tomorrow
19mSupercharging Kafka Applications with Real-time Contextual Insights