When you want to build a geographical distant architecture, you need to think about several things:
Should I replicate everything?
Is write speed important?
What is the expected read load for each site?
The following figure shows an example with a French Datacenter where we've got the master node and several slaves that distribute information to other countries (Brazil and Japan):
The latency and bandwidth with such distances are problematic for replication. Consider the following important information:
Should I replicate everything: In the previous section, we saw how to replicate only some databases or tables and how to exclude some of them from replication. It's important to understand and know the purpose of the replication. Is this done for a site extension or is it for a full replication such as disaster recovery? If you want to build a full replication where you can completely start to recover, then you need to replicate everything! But, you should...