How a weather app serves a forecast for any point on earth | system design

Nobody computes a forecast per request. They precompute a grid and interpolate.

Forecast models run on a fixed grid refreshed every few hours, and your exact spot is interpolated from the nearest grid points, then cached at the edge.

Weather changes constantly and people ask for arbitrary points, so computing fresh per location, per request, doesn't scale.

How it's fast:
- Forecast models run on a fixed grid of points, refreshed every few hours
- Your exact spot is interpolated from the nearest grid points, which is cheap
- Cache results at the CDN edge by region, since neighbors share nearly the same forecast

The catch: a coarse grid is cheap but blurs sharp local weather, like a storm cell sitting between grid points. A finer grid costs far more compute. You pick resolution by how much accuracy is worth.

Coarse grid that's cheap, or fine grid that catches local storms?

#systemdesign #techinterview #softwareengineering #codinginterview #faang #interviewprep #weather #geospatial #cdn #caching #backend

Видео How a weather app serves a forecast for any point on earth | system design канала The System Round