Skip to content
This repository was archived by the owner on Mar 11, 2024. It is now read-only.
This repository was archived by the owner on Mar 11, 2024. It is now read-only.

Low performance when using EventTimeInputBuilder #65

@fschoell

Description

@fschoell

I'm currently using flink-spector to read a "large" (about 10.000 values) dataset from a CSV file and use the EventTimeInputBuilder to create a data stream from it. However, the performance drops significantly the larger the input gets. My application needs about 30 minutes to process those data (even though the algorithm basically just calculates a moving average). Cutting down the input to 5.000 values drops the runtime to 5 minutes.

I used a profiler to dig into the problem and it seems that InputUtil.produceWatermarks is quite time consuming.

I'm using the InputBuilder now and assign timestamps and watermarks myself which drops the runtime to a couple of seconds. Just wanted to report this.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions