Hierarchy processing too slow

Please refer to Case TS000137042 for details.

As IBM already knows MunichRe is having major issues with the performance of the TM1 system, especially with the performance of hierarchies (in reporting as well as in processing).

This ticket is for requesting support to achieve major performance improvements while creating or changing hierarchies. Christian Wagner promised that it should be able to change even big hierarchies within seconds. This is clearly not the case. The attached process proves this. The following findings can be reproduced with this process:

- Duration to create a dimension and hierarchy with 4 million elements takes roughly 10 minutes on our system.
- Adding a single element to this dimension and hierarchy takes 2 minutes!
- After adding elements to the hierarchy the number of elements in the Leaves hierarchy is incorrect.

How to reproduce:
- Start the provided process without changing any parameters. After 10 minutes you should have a dimension and hierarchy with 4 million elements (in case your test system is as powerful as ours).
- Start the process again with the parameter "action" set to create. Now it takes two minutes to add one item to the hierarchy. Two items! This is our major pain point with respect to hierarchy processing! We need to be really a lot faster with that (few seconds would be acceptable). In the production environment we need to change 30 hierarchies 3 times a day. While this process is running the system cannot be used anymore. And we can't run this "in the night" as we need to serve users all over the world.

And as we understood from Christian it is not the expected performance from IBM as well we need to get better on this. IBM was informed from the very beginning that MunichRe heavily relies on the feature of hierarchies and its use in the project.

So we request the following:
- Let us know whether there is anything wrong in the coding of the sample process. If somehow possible provide the process back to us with expected performance of few sec.
- If there is no coding issue with our sample please provide a fix for TM1 to behave as expected.
- Let us know what is TM1 doing between Prolog and Epilog. It takes 5 minutes from the end of the prolog to the beginning of the epilog on our system while creating the hierarchy. It would be a major performance improvement if this time would not be lost!
- If you run above given process with only 1 million elements you can see that the performance is increasing faster with less elements than expected (I could understand a linear scaling but it is worth).

In case you need more information please let us know. But the provided process should basically log all relevant information.