- IMDC does not keep track of partitioning.
An IMDC partition is associated with each parallel instance of Aster Database. A parallel instance contains whatever data you load into it. If you load replicated data into n parallel instances, and then project the data from the IMDC, the output is n instances of the replicated data.
- A SQL-MapReduce function that uses an iterator to consume data might not be invoked for empty partitions.
The recommended workaround is to implement the Drainable interface and in the drainOutputRows method, check for missing IMDCs. If an IMDC is missing, create an empty IMDC and finalize it. See Finalization Code for a Partition Function.