Jack Gudenkauf - HP Big Data 2015 - theCUBE
summary
Playtika runs a pipeline called PSTL pipeline that ingests every event — every slot spin, point change and payment — and updates game versions weekly to drive real-time changes. Data streams through Kafka → Spark → HDFS → Vertica where Spark does in-memory transforms, HDFS holds cold storage, and Vertica accepts parallel loads for analytics. The architecture shortens a week-long update loop toward near real-time predictive adjustments across ~10 game properties and ~1,000 employees worldwide.
AI-written summary · about
chapters (16)
- 0:00Introduction to HP Big Data 2015
- 0:45Speaker introduction and background
- 1:26Data's growing role in modern apps
- 2:31Gaming industry's data-driven evolution
- 3:54Customer-driven game development
- 4:58Social gaming features and integrations
- 6:23Introducing the PSTL architecture
- 7:26Spark's practical advantages over Hadoop
- 9:57Hadoop and Spark coexistence debate
- 12:47Stonebraker's insights on architecture
- 14:51Kafka, Spark, and Hadoop pipeline
- 15:54Analytics evolution and predictive models
- 17:39Game economy and user behavior
- 18:22Parallels between gaming and real-world systems
- 19:06Why HP Big Data event stands out
- 20:10HP's engineering roots and culture
