Climbing the data mountain: Scaling and optimizing the delivery of vector tiles
At mySidewalk, performance isn't just a feature, it's THE feature. Over the last year, we've spent most of our time trying to make the visualization and analysis of geospatial data faster and easier, and vector tiles have been the special sauce that has allowed us to take data delivery and visualization to a whole new level.
In this talk, I will discuss several performance optimizations that are implemented at various steps in the vector tile pipeline: from client requests, to the tile server, the database, tile caching, and back. Specific technologies/strategies to be discussed are:
- requesting and rendering tiles using Leaflet
- reverse proxies and load balancing
- tile generation and delivery with TileStache and PostGIS
- database optimization with PostgreSQL
- cache optimization and management with Redis
- infrastructure and scaling with Docker
Generating and delivering vector tiles on the fly and at scale isn't a simple task, but with some planning and elbow grease, it's definitely attainable. Whether you're currently generating vector tiles, or would like to in the future, this talk should prove to be valuable in helping you build a solution that will perform for your users.