In this type of topology, users can communicate with any data center. There is no single point of failure, and any element can be removed or fail transparently.
The following illustration shows an example of multi-datacenter clustering. It includes two data centers that each have a database, a file share, and two servers. The file shares and databases are synchronized via replication.
Changes to one database are replicated in the others. There are many techniques for replicating data, such as asynchronous, synchronous replication, or transactional.
Master-master or multi-homing data centers as described here can work well if there is a low latency connection between data centers.
The following illustration shows one viable approach.