The closest publicly available problem I can think of is the 1 billion rows challenge. It's got a bigger dataset, but with somewhat simpler statistics – though the core engineering challenges are very similar.

https://github.com/gunnarmorling/1brc