If you were to rebuild 4chan today...

...what type of data storage layout would you use?

Would you remain with a mostly flat file layout and what technologies would you use to implement such a layout?

Personally I'd go with JSON files and some random JSON library for writing, cache and reading stuff.

>>5
According to the front page there's 92GB of content in total for all of the boards.

If they split the boards across different servers they could hold everything in memory and not bother with persistent storage at all.

In that case, just store all the HTML pages and images in a huge in-memory hashtable.

128GB of DDR3 can be had for <$1000.