It's mostly metadata they store end up storing. So who is talking to whom at what time using what protocol and certain other details. We know from Snowden leaks for instance that they were doing image captures every 6-10 seconds of webcam sessions via things like Yahoo messenger and that they were shocked that so many people used them for sex and issued a warning to employees not to abuse access to those resources. So it's not storing absolutely everything just key details.

hmmmm, today I will wonder what a distributed computer system is. What is Apache? Must be firewall tech that I can still access Chinese media/websites from.