Keynote: How Can We Take Flink Forward?
summary
Flink is already a unified streaming-and-batch engine with technical advantages, but the pivotal need is production adoption not more features. A prototype for a financial client uses topic-keyed queues and MapR Streams to eliminate the database, run on 4–5 nodes and sustain millions of transactions per second by combining persistence with extreme performance. Interop efforts (Calcite, Beam, Kafka, Drill, Arrow) and solving multi-tenancy/security are the concrete barriers to wide deployment.
AI-written summary · about
chapters (13)
- 0:00Introduction and welcome
- 1:27Current state of Flink
- 2:08Digital revolution and data value
- 8:10Big data superiority over algorithms
- 10:38Microservices and vertical divisions
- 13:28Streaming idiom and deferred work
- 16:20Transport and streaming computation
- 17:45Physical limits of computation
- 20:13Flow-based computing revolution
- 22:41Flink’s role in future software
- 23:23Business impact of real-time data
- 24:28Flink’s future and adoption challenges
- 26:15Audience discussion and Q&A
related talks
40mInteractive Session - Live Code Contribution to Apache Flink
09mSponsored Keynote: Building a Financial Lakehouse with Legend
39mKeynote: CNCF Project Updates
41mOpen Source Social Applications with Apache Streams
21mCase Study: FOSS (Apache Hbase) at Flipkart
42mIngestion at Scale: Open Source Pipelines for Getting Data Into Open...
18mKeynote: CNCF Project Updates
17mKeynote: Global Industry, Global Innovation: How FINOS is Delivering Value to...