Design Real-Time Trending Likes

Design a real-time analytics service for a large social platform. The service receives like and unlike events for posts. It must support the following queries: 1. Given a `post_id`, return the number of likes that post received in the last 60 seconds. 2. Return the global top `K` posts by number of likes received in the last 60 seconds. 3. Return the top `K` posts by number of likes received in the last 60 seconds for a given region. Assume the platform may receive up to about 1 million like events per second at peak. The system should have low read latency and tolerate high write throughput. Clarify whether the 60-second window is a sliding window or tumbling window, whether counts must be exact or approximate, how regions are defined, and whether unlikes should decrement the counts.