You can also have a system with multiple master servers. If there are two servers, you have a master-master replication scheme. If one of the servers dies, the others can still operate normally. However, now you either need to synchronize the writes or provide looser consistency guarantees. Also, you need to provide a load balancer.
Examples of such replication include Microsoft's Active Directory, OpenLDAP, Apache's CouchDB, or Postgres-XL.
Let's now discuss two ways to prevent faults caused by too high a load.