No deduplication...
That is probably why they have 6,000TB of swag. If I were to run a Usenet server, or file sharing site, or whatever, I'd definitely invest in deduplication -- either in hardware or software. Because, honestly, what they probably have is like 50 different people putting up the new..umm... Linux distro... yeah. And 35 copies of the other one. And so on.
For those who don't know, deduplication compares new data to existing data (either at the whole-file or the block level), and if the data is identical, only physically stores one copy. (Well, hopefully at least 2 copies, in case of a disk failure.) The upside? Potentially massive reduction in storage used. The downside? All that extra CPU time, disk I/O, and housekeeping to determine what can be deduplicated. Deduplication is either done real-time (reducing write speeds, perhaps considerably) or as a batch process (so you get essentially full speeds when you need it, and queue up the deduplication to happen at 2AM or whatever when your disks are at minimal load anyway.)