What Is Data Replication?
In its classic definition, data replication is based on the ”store and forward” data distribution model, as shown in Figure 7.2. Data that is stored in one location (inserted) is automatically “forwarded” to one or more distributed locations.
Figure 7.2. Store and forward distribution model.
Of course, a more complete data distribution model would address updates, deletes, data latency, and autonomy for data and schemas. It is this data distribution model that Microsoft's data replication facility serves to implement. There are three primary scenarios for using data replication:
The first, as you can see in Figure 7.3, is for when you need to offload the processing from the primary server for reporting purposes to a separate reporting server. This server is typically in read-only mode, and contains all the data that is needed to satisfy report requirements.
Another is for when you need to provide vertical or horizontal subsets of the primary server's data to other servers so that they are working (updating) only their own data. The primary server should contain everyone's data, though. This can directly support high availability. This is common when you need to have your European and Asian users using their own data but corporate headquarters is continuously fed their changes.
And finally, you could be replicating all data on a server to another server so that if the primary server crashes, users can switch to this fail-over server quickly and continue to work with little downtime or data loss.
Each one of these scenarios supports parts (or all) of a high availability solution.
You can use data replication for many reasons. First, however, you need to understand some of the common terms and metaphors used by Microsoft in data replication (as will be described in this chapter).