Apache Flink Meets Apache Mesos and DC/OS
summary
Explains a concrete Mesos/DC/OS integration: Flink becomes a Mesos framework using Fenzo offer matching and a new FLIP-6 architecture that adds a long‑running dispatcher and per‑job resource bundles, enabling dynamic scale-up and scale-down of TaskManagers while jobs run. Demo wires Kafka → Flink → Kafka to flag money‑laundering when transactions sum > $10,000 using Flink event‑time windows; state checkpoints land on HDFS so workers can be swapped for rolling or blue/green upgrades. Packaging targets DC/OS for easy install.
AI-written summary · about
chapters (26)
- 0:00Introduction
- 3:51Use Cases
- 5:32MAC Stack
- 6:21Stream Processing
- 7:10Spark
- 7:45Spark Streaming
- 8:53Spark Execution Model
- 9:37Flink Processing
- 10:13Flink Fault Tolerance
- 11:20Flink Delivery Guarantees
- 12:34What does it mean for our data centers
- 13:12The vision of Mesos
- 13:37Challenges of Mesos
- 14:25Cassandra
- 15:08Flink SDK
- 15:33Flink Scheduler
- 15:48Flink JIRA
- 16:15Why Flink
- 17:20scheduler parts
- 17:58resource manager
- 20:42Flink library
- 21:28New Flink architecture
- 23:09Demo
- 34:02Cluster Utilization
- 35:02Conclusion
- 35:42Questions
related talks
41mIntroducing Flink on Mesos
41mRunning Apache Flink everywhere
20mGetting Started with Event Stream Processing via Apache Flink on Microsoft Azure
40mFrom concept to production; the fastest way to develop & deploy Apache Flink jobs
36mApache Flink in the Cloud-native Era
26mRunning Apache Flink on AWS
44mMulti Cloud Mesos
34mBeaming Flink to the Cloud @ Netflix
