Key server synchronize using different protocols, forming a network of more than hundred servers all around the world involved.
The "old" way of key server synchronization is based on the key servers sending e-mails to each other. Whenever a key server receives new information he doesn't know yet (either uploaded by a user or received from another key server), he forwards these information to all other key servers in his synchronization list.
Because the old e-mail based protocol doesn't scale very well (especially with the number of servers in the synchronization list), a new protocol was defined for the Synchronizing Key Server (SKS), which is based on set reconciliation.
These key servers "gossip" with each other in a given time interval. From a very basic point of view, set reconciliation orders the keys in a so-called partition tree, which allows to find the differences easily without transmitting a lot of information. Only the modified keys are then exchanged.
MIT and University of Mainz Key Servers
For SKS keyservers (and also Hockeypuck, which is another implementation of the SKS protocol), the gossip partners can be retrieved by fetching their statistics page, available as
http://[keyserver]:11371/pks/lookup?op=stats. Looking at the statistic pages of both the MIT key server and the one of the University of Mainz, one will realize they even have a direct, mutual synchronization agreement (look at the "Gossip Peers" column).
The key server of the CCC Hanau on the other does not directly gossip with the one of the MIT. If you upload a key to any of these servers, the other will not get aware of it directly, but through the intermediate server of the University of Mainz (or on any other synchronization path).
Visualizing the Whole Key Server Network
The SKS keyserver pool crawls the key server network periodically, and provides a dot file of the key server network (of all servers in the SKS pool, not including synchronization links based on the old e-mail synchronization algorithm).