To set up a distributed system, you need to divide the source system data into data packages using a suitable criterion. One data package is associated with one sub-server. Please ensure that associated process instance data need to be assigned to the same sub-server. If you are using process hierarchies, make sure that the sub-servers each contain the complete tree branches of the hierarchy.
Which criterion you select depends on the use case. For example, the process of dividing the data packages could include dimensions such as the location: Sub-server 1 receives the data from location A, sub-server 2 the data from location B, etc. The Point in time criterion is rather unsuitable, though, because it usually limits parallelity of analysis queries and imports.
The data packages are imported directly into the relevant sub-servers. You can select the time of import for each sub-server. Import can, for example, be executed on multiple servers simultaneously in order to reach maximum parallelity.