A couple decades ago I downloaded lists of given names and surnames and their frequency from the US Census. I used it to randomly generate names for populating an LDAP server for benchmarking, to have a more realistic data set than just "user xxxx". I also used lists of street and city names, with random ZIP codes.

Haven't used that in a long time; the workloads we see nowadays don't tend to have postal addresses. Funny how data relevance changes over time.