First Unique IP Hitting the Server (Streaming)

Problem: First Unique IP Hitting the Server (Streaming) Your server receives a continuous stream of IP hit events. Design a data structure that supports: add(ip): record one hit from ip. first_unique(): return the earliest IP that has appeared exactly once so far; return empty (e.g., null / empty string) if none exists. Requirements Make both add and first_unique as efficient as possible. IPs are given as strings (e.g., "10.0.0.1"); you do not need to validate the format. Example Calls: add("1.1.1.1") add("2.2.2.2") add("1.1.1.1") add("3.3.3.3") Now first_unique() should return "2.2.2.2" (it appeared earliest among those that appear exactly once). Constraints Number of events N up to 10^7. Number of distinct IPs U up to 10^6. Follow-up: Scalability (Millions of Records) When the data volume is huge (e.g., tens of millions of events, millions of distinct IPs, and potentially multiple writers across machines/data centers): How would you scale the design without significantly increasing query latency? Explain your sharding/storage/consistency/memory-control approach (no code required). Example Input add 1.1.1.1 add 2.2.2.2 add 1.1.1.1 first_unique Output 2.2.2.2