Flink Table Store: Streaming data warehouse architecture and scenario

李劲松

Chinese Session 2022-07-31 14:50 GMT+8  (ROOM : A) #bigdata

Flink Table Store is a unified streaming batch storage for streaming data warehouse. It is used to build dynamic tables for streaming batch processing in Flink, supporting real-time streaming consumption and real-time OLAP query. The Flink Table Store has released its first forward-looking version, but is missing a lot of work on ecology and stability. We are currently working on the second version, and we hope that the second version will bring more production capacity. Through this sharing, you can see what we are enhancing to improve the reliability and ecology of storage. In addition, I will share the subsequent architecture, the Service version, how it achieves a unified stream store, and what scenarios it unlocks.

Speakers:


JingSong Li: alibaba, Technical experts, Currently, I am working in Aliyun open source Big data, engaged in the field of distributed stream/batch processing system for a long time, and also have some research on data lake and OLAP MPP. I am an Apache Beam/Flink/Iceberg Committer, and have a certain understanding of the underlying scheduling, communication mechanism, user model, SQL stream batch computing, and storage. At present, I focus on the development of Flink Table Store project, hoping to bring a most suitable storage to Flink.