Question
Design the purge subsystem for a CDN serving a news site during a breaking-news event. An editor corrects a headline and the change must propagate to ~250 edge PoPs (each with thousands of cache nodes) in under 5 seconds, but a single article correction can touch tens of thousands of cached objects (the article, every section page that embeds it, AMP variants, RSS, API responses). You see purge bursts of 100k+ objects in a few seconds during big stories. Describe the purge data model, how a single editorial action maps to the right set of objects, and how you make global purge both fast and resilient to PoP partitions.
Clarify scale and constraints first. Propose a clean component breakdown, then go deep on the hard parts — data model, bottlenecks, consistency, failure modes — and name the trade-offs you are making.