Heaps & Intervals

Cloudflare fronts a large chunk of the web from an edge network of hundreds of points of presence, the data centers it calls PoPs. Each one terminates connections, serves CDN cache, and absorbs attack traffic close to the user. A lot of the work there is counting and windowing. Which source addresses are sending the most requests right now? Which time or address ranges overlap? How many connections are live at once? Two patterns cover most of that. A heap hands you the extreme of a set that keeps changing without ordering the rest. A sweep walks a sorted timeline of intervals to see where they pile up. This family is a smaller slice of the coding loop than hash-map design or plain arrays. But it maps so directly onto edge telemetry that you should have it ready.

The problems here are canonical, asked close to the way you already know them. The pressure Cloudflare applies is the one it applies across the coding rounds. The code has to run and pass the cases on the shared pad, whether that is CoderPad or HackerRank, and you have to say the time and space cost out loud. A solution that only clears the happy path is not enough. The interviewer almost always has a follow-up waiting. So a clean size-k heap and a clean interval sweep, written without fumbling, buy you room to handle what comes next. Usually that is a scaling twist on the same algorithm.

The Cloudflare angle

A PoP under load needs to know which source addresses are hammering it hardest, and that ranking is a top-K heap. Shaping a DDoS wave or flagging the noisiest client both come down to one question: of everything in these logs, which few keys have the highest count? That is the top-K cut behind Top K Frequent Elements. Count per source, then keep only the k largest in a size-k min-heap, evicting the weakest as stronger ones arrive.