Find Global and Rolling Top-K URLs Under a Memory Limit

You are designing a URL analytics service that consumes a high-volume stream of access events. Each event contains at least a canonicalized URL and an event timestamp. The service must support two types of top-K queries: All-time top K URLs by total access count over the entire retained history. Rolling top K URLs by total access count over the trailing 24 hours, measured using event time. The number of distinct URLs may exceed the memory available on a single host. Describe…