Open
Description
Describe the feature
In enterprise production practice, if the built-in functions of the computing engine cannot meet user needs, users usually customize UDFs(User-Defined-Functions) through relevant interfaces and reference them in the engine. Especially for scenarios that require a lot of reuse, it is very important to effectively manage these custom functions. In Xiaomi's internal practice, there is also a UDF management system that basically meets user needs (limited to Spark/Flink, and there is no good support for AI + Python scenarios).
As a unified Data / AI metadata center, I think Gravitino can also support UDFs management in Data / AI scenarios.
The following are some commercial product practices:
- Aliyun: https://help.aliyun.com/zh/flink/user-guide/manage-udfs
- UnityCatalog: https://docs.databricks.com/en/udf/unity-catalog.html
Motivation
No response
Describe the solution
No response
Additional context
No response
Activity