doris VS flink-kubernetes-operator

As an open-source real-time data warehouse, Apache Doris provides semi-structured data processing capabilities, and the newly-released version 2.1.0 makes a stride in this direction. Before V2.1, Apache Doris stores semi-structured data as JSON files. However, during query execution, the real-time parsing of JSON data leads to high CPU and I/O consumption in addition to high query latency, especially when the dataset is huge and complicated. Moreover, the lack of a pre-defined schema means there is no handle for query optimization.

Apache Doris is a real-time data warehouse.

Learn more about Apache Doris or find the Doris makers on Slack.

As you can imagine, a long and complicated data pipeline is high-maintenance and detrimental to development efficiency. Moreover, they are not capable of ad-hoc queries. So as an upgrade to our data warehouse, we replaced most of these components with Apache Doris, a unified analytic database.

GitHub source code: https://github.com/apache/doris/tree/branch-2.0

The key to Architecture 3.0 is the combination of Flink and Doris, so this is how to connect them. Probably the most important code in building architecture 3. flink-demo stream-load-demo

https://github.com/apache/doris Is a great example. Same for it's cousin https://github.com/StarRocks/starrocks that was an early fork of the doris project.

To be fair, these are the only examples I can think of and I only learned of these as I'm standing up new data infra using starrocks.

We optimized Apache Doris to solve these problems. (Pull Request on Github)

So, in Storage Architecture 2.0, we introduced Apache Doris and Apache Spark. The whole data pipeline was a Y-shaped diagram.

Data scientists often prefer Python for its simplicity and powerful libraries like Pandas or SciPy. However, many real-time data processing tools are Java-based. Take the example of Kafka, Flink, or Spark streaming. While these tools have their Python API/wrapper libraries, they introduce increased latency, and data scientists need to manage dependencies for both Python and JVM environments. For example, implementing a real-time anomaly detection model in Kafka Streams would require translating Python code into Java, slowing down pipeline performance, and requiring a complex initial setup.

Other stream processing engines (such as Flink and Spark Streaming) provide SQL interfaces too, but the key difference is a streaming database has its storage. Stream processing engines require a dedicated database to store input and output data. On the other hand, streaming databases utilize cloud-native storage to maintain materialized views and states, allowing data replication and independent storage scaling.

also, this knowledge applies to learning more about data engineering, as this field of software engineering relies heavily on the event-driven approach via tools like Spark, Flink, Kafka, etc.

Apache SeaTunnel is a data integration platform that offers the three pillars of data pipelines: sources, transforms, and sinks. It offers an abstract API over three possible engines: the Zeta engine from SeaTunnel or a wrapper around Apache Spark or Apache Flink. Be careful, as each engine comes with its own set of features.

Due to the technology transformation we want to do recently, we started to investigate Apache Iceberg. In addition, the data processing engine we use in house is Apache Flink, so it's only fair to look for an experimental environment that integrates Flink and Iceberg.