[EPIC] GeoIO Acceleration

[jarmak-nv]

Overview and Description

The goal of this epic is to enhance cuSpatial's performance and efficiency by implementing IO acceleration for various geospatial dataset formats. As the volume and complexity of geospatial data continue to grow, the ability to quickly read, write, and process this information becomes increasingly important for users across diverse industries and applications.

cuSpatial's GeoIO Acceleration will minimize bottlenecks, reduce data processing times, and deliver a more responsive user experience. This effort will involve optimizing file parsing, data loading, and data writing operations for popular geospatial formats - specific formats to be determined prior to the 23.08 release.

The outcome of this epic will be a more robust and efficient cuSpatial, better equipped to handle the demands of modern geospatial data processing and analysis.

Target start: During the 23.08 development cycle Target release: Release 23.10

Requirements

For successful completion, this epic needs:

- [ ] Discussion and decision on which geo formats to accelerate
- [ ] Discussion and decision on UX when considering formats we do support, and those we do not
- [ ] libcuspatial implementations
- [ ] Python bindings

Considerations

Today, all IO operations are expected to be through geoPandas. Through pyogrio many data formats can be read straight into geoarrow. cuSpatial, also representing data in the geoarrow format, should explore how this can be leveraged.

Related Issues

This epic makes progress on, and potentially may close:

• https://github.com/rapidsai/cuspatial/issues/630
• https://github.com/rapidsai/cuspatial/issues/601