Skip to content

WindowedArrays not performing as expected when streaming data from copernicusmarine #2756

Description

@erikvansebille

Parcels version

v4

Description

As described in Parcels-code/parcels-benchmarks#49, using to_windowed_arrays() when streaming data directly from the copernicusmarine store is approximately 25x slower than temporarily storing the files to disk.

As streaming the data from CopernicusMarine is the preferred way (no need for temporary storage, and able to scale to much larger datasets than required here), this slowdown is not acceptable.

Code sample

See the script at https://github.com/Parcels-code/parcels-benchmarks/blob/Copernicus_streaming/benchmarks/copernicusstreaming.py

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugperformanceIssues to do with performance

    Type

    No type

    Fields

    No fields configured for issues without a type.

    Projects

    Status
    Backlog

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions