DataLoader/MX Reference Manual (H06.03+, J06.03+)

Recovery Strategies
DataLoader/MX Reference Manual—543544-001
7-9
Parallel Considerations
•
Have the downstream DataLoader/MX processes read from the upstream
DataLoader/MX process. This method is self-balancing because each process can
get another block of records as soon as it has finished its current block of records.
Downstream processes that are running fast because they are in processors with
little competition can read as many records as they are able to and are not held up
by other downstream DataLoader/MX processes that are running in a busy
processor.
•
Have the upstream DataLoader/MX process write records to the downstream
DataLoader/MX processes. This method is not self-balancing. A slow downstream
DataLoader/MX process can hold up distribution of records to all the other
downstream DataLoader/MX processes. However, this method can be restarted if
you use the method of restarting from a known state by using the -T parameter.
For a loading job to be restarted using restarting from a known state by using the -T
parameter, each DataLoader/MX process must receive the same records and in the
same order as were read in the first run. By using the scenario in which downstream
DataLoader/MX processes read from the upstream DataLoader/MX as fast as they are
able to be read, the records that a given downstream DataLoader/MX receives are
controlled by the system’s load, and their order cannot be repeated exactly during a
rerun.
If you are using restarting from an unknown state and are not using -T, the records that
a given downstream DataLoader/MX receives are not important. It is an advantage to
restart from an unknown state because you can use a self-balancing parallel
configuration.