In April I presented Addresscloud's Intelligence service at the 2021 GIS Research UK (GISRUK) conference. The presentation brought together our work on different components of the API as an example of a complete geospatial application built using a serverless microservices architecture.
In previous talks I've detailed our experience building serverless data-stores for raster and vector data, and how we use the H3 geospatial index to cache data within a NoSQL database. In this presentation, now available to watch on YouTube, I discuss how our Intel service architecture combines all three of these components to respond to thousands of customer requests simultaneously - across multiple data-types - without increasing system latency. In addition to the video I've also included a short summary of key-findings and the full conference paper below for those interested.
A big thanks to the GISRUK team at Cardiff University for hosting such a well organised and engaging conference.
- Addresscloud Intel is an example of a serverless microservices architecture linking multiple scaleable data-stores to serve geographic queries.
- H3 Spatial Index is used to cache property data using a low-latency NoSQL database.
- Serverless PostGIS and Cloud Optimised GeoTiffs are used to serve requests not covered by the cache.
- Design enables consistent performance with 99.9% of queries resolved within 500 milliseconds.