A distribution is a set of table mappings designed to be moved as a unit by a request. You must create one or more distributions to contain your table mappings, and define one or more requests to specify the plan for moving the distributions.
In a simple broadcast to multiple targets, where conditions are the same for each target, a single distribution might contain all the target table mappings. However, if a subset of a distribution requires a different data movement request, the subset must be defined as a separate distribution.
In this example, two distributions and two copy requests are necessary because at one of the target sites, the source data is to be copied and merged with the existing target data instead of entirely replacing it, which is the (default) mode used by the other table mappings. At this site, the source data is divided between two audiences, each of which accesses a table on a different server.
Consequently, the table mappings in this example that require a simple replace and copy request with a Refresh Target Apply Mode are defined as one distribution, and the table mappings requiring a different copy request are defined as another distribution.