The basic idea is to count the number of unique users. Since the complexity of the problem grows with the users maintaining a count of unique users becomes a space issues.
Apart from what’s written in the blog I would suggest having a look at the hyeprloglog algo or having a bloom filer.
If we want to maintain a list of users who have been active at what time – in that case we can make buckets for each second or one bucket for each minutes and store the user ids corresponding to the bucket.
Read this as well – http://stackoverflow.com/questions/17562089/how-to-count-number-of-requests-in-last-second-minute-and-hour