There are various data replication models for replicating data.
Basic primary copy model – centralized primary data, distributed replicate data
Distributed primary fragments model – both primary and replicate data distributed through the replication system
Corporate rollup – distributed primary data, centralized replicate data
Redistributed corporate rollup – same as corporate rollup, but updates redistributed to replicate databases
Warm standby applications– two databases, one serving as backup for the other, which together as a logical unit may participate in replication
The type of application you are building, the way you update primary data, and the way you manage potential update conflicts determine the model you use to implement your replication application.
For instance, you can use the basic primary copy model to implement either a decision-support application or a low-volume distributed OLTP system. You might implement a decision-support application using either the basic primary copy model or the redistributed corporate rollup model, depending on whether your primary data is centralized or fragmented. A distributed OLTP application might be implemented using the distributed primary fragment model, with or without corporate rollup, depending on additional decision-support needs.