The use of cloud technologies such as those available via Amazon Web Services (AWS) opens up opportunities to rethink how we ingest, manage and disseminate geospatial data using open standards. AWS provides a platform on which to build efficient and scalable solutions which utilise serverless architectures.
This presentation will provide an overview of a recent project undertaken by the Australian Geospatial-Intelligence Organisation (in coordination with Geoplex) to emulate a WMTS API which dynamically generates and delivers raster tiles in response to HTTP/S requests. Topics covered will include:
- Architectural design overview;
- Preprocessing and indexation of a very large (187TB) JPEG2000 imagery dataset within the AWS environment using GPU clustering and spot pricing;
- Use of GDAL and Python to generate raster tiles on-the-fly within the AWS Lambda serverless compute framework;
- HTTP/S request/response handling using AWS API Gateway, including request transformation;
- Strategies for performance optimisation, including multi-level caching and optimised index queries;
- General reflections on the implementation and consideration of the future direction of the project.