You must map specific process criteria to particular columns of the process data table to specify individual processes. That is, a row of the process data table defines a function with function name and process ID, among other things. And all functions with the same process ID belong to the same process instance. The sort criterion defines the order of the functions within the process instance.
Mandatory process criteria
Process ID is of TEXT type, and mapped to AT_ID on process level.
Function name is of TEXT type, and mapped to AT_OBJNAME and AT_OBJNAME_INTERN on function level. Note that if you want to define own calculated attributes, the function names must be conform to XML NMTOKEN. Otherwise, it would not be possible to filter a specific function type with "filteredattribute".
Sort criterion is of TIME type, and mapped to AT_SORT_MERGE_TIME on function level. The sort merge key can be mapped to either function start time or function end time.
Optional process criteria
Function ID is of TEXT type, and mapped to AT_ID on function level.
Function start time or function end time is of TIME type, and mapped to AT_START_TIME or AT_END_TIME at the function level.