medium10 pts · first answer countskafkastream-processing

Unique visitors per minute at scale

A stream job must report unique visitors per minute for a site with 50 million users, and a 1–2% error is fine. What keeps state small?

A.A HyperLogLog sketch per windowB.A set of all user IDs per windowC.Sorting user IDs and counting changesD.Counting every event as a new user

Discussion 0

Join free to post in the discussion.