Skip to content

Parallelize Serialization of Trino Data to RecordBatches #7

@devinjdangelo

Description

@devinjdangelo

Is your feature request related to a problem or challenge?

Serializing trino json-like data record to RecordBatches is likely to be a performance bottleneck.

Describe the solution you'd like

Parallelizing serialization on the Relay so the trino client continues to fetch data while also serializing on multiple cores.

This should be straightforward to tokio spawn serialization tasks, but the complexity may be in creating a RecordBatchStream from an mpsc async receiver of the parallel tasks.

Describe alternatives you've considered

Don't parallelize serialization as network IO may bottleneck anyway, though I find this unlikely if trino connector were to go up against FlightSQL in the same query.

Additional context

No response

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions