How to build a redundant cluster of servers (From my blog): http://tomoconnor.eu/blogish/dedicated-dedicated/
It's really difficult to give specific models and manufacturers, because that kind of information is really specific and localised. You'll want 2 or more of everything, LOTS of memory, FAST disks.
Anything by Dell or HP that fits that category will be fine.
Bare in mind that Mongo uses lots of in-memory space for data, so you'll want a server that can take craploads of DIMMs. Perhaps an HP Bladesystem might be a better fit than 1U rack servers.
==edit==
3-way replication, and 11TB+ of database.
That is an enormous amount of data. Bear in mind that Mongo stores/caches a lot of data in RAM. Assuming you'll want some kind of speed, you should definitely be looking at some very big servers. Earlier on we were discussing this in The Comms Room and came to the conclusion that if you wanted to store as much of it in memory, then you'd have to shard it very heavily, and have multiple shard servers, to get your redundancy goals.
This is not going to be a cheap project.