Databases · Senior · Question 36 of 75

How can you count unique visitors over huge data sets cheaply when approximate results are fine?

1Store every visitor ID in a list and count it
2Use COUNT(*) on the raw table every time
✓Correct answer: Use a HyperLogLog sketch (e.g. Redis PFADD/PFCOUNT), which estimates distinct counts with tiny memory

Explanation

HyperLogLog keeps about 12 KB per counter in Redis with around 1% error. Sketches can also be merged across days or shards.

Further reading

Test yourself

Take the Databases · Senior quiz: questions and options are shuffled, and you get a review of your mistakes at the end.