Today’s topic is about deleting or adding dimensions to TM1 cubes that you are using in Production. With TM1’s introduction of Performance Modeler, you now have the ability to add or delete a dimension from an existing cube. Before this, you would have had to create a whole new cube. While this is pretty simple when done in the build phase, there are several dependencies to be mindful of, especially if you do this on a cube that is currently in use. It’s a nice feature to be able to do this, but it does open up Pandora’s Box…
Dependent Cubes: Utilize the Model Flow Chart functionality in Performance Modeler so that you can see what other cubes are dependent on the cube you are changing. This will show you the dependencies of rules, feeder and links to other cubes that may be affected. Unfortunately, this will NOT let you know what TI Processes are affected.
Cube Dimension Order: Take note of the dimension order of your cube BEFORE you make your change. Since you are using Performance Modeler to add/delete a dimension, it will be the driver of the dimension order. So for example your cube was originally ordered with the dimensions of Year, Month, Measure and you decide to add Version. TM1 may change it to Month, Version, Year, Measure. Be mindful of this, because the correct dimension order is necessary in scripting of any rules that use a DB function and in any TI processes.
Views: Since you are changing the structure of the cube, views (and any private views) will become invalid. Take inventory of existing views and how they are set up for both dimension position and subsets used by those dimensions. This is especially important if views are used in TM1 Web and TM1 Applications (aka Contributor). You will have to recreate the views and if using TM1 Applications, update your Application and redeploy it. The Application section of Server Explorer for views used in TM1 Web will need to be updated too.
Existing Data: You will be prompted how you want to handle your existing data. The choices are Clear Cube Data, Duplicate Data or to specify an element from the dimension changed.
Rules/Links/Cube Calulations: Any manual rules and feeders that use the DB function will need to be updated. Make sure you have your dimension order correct in the DB functions when you update them. Links and Cube Calculations may require updates too.
Websheets: Any websheets that are dependent on this restructured cube will need to be updated. Websheets pull in data by the coordinate of the cube cell, so again, you need to update the formulas with the correct dimension order. Communicate this with your users before you make your intended change. You may have some power users that have their own websheets that you don’t know about and this will affect them too.
TI Processes: These will also need to be updated for whatever the data source is – even if it is a process using the same cube.
Cognos BI, Cognos Insight & 3rd Party Software: You will have to be cognizant on what is used here as well.
In the end, there are valid reasons for deleting or adding dimensions to TM1 cubes, so I am not advocating against it. It’s quite similar as if you created a whole new cube to replace an existing cube. You just need to think through all the ripple effects. For more information, don't hesitate to contact us at Sales@LodestarSolutions.com or check out this link for the IBM Performance Modeler user guide: http://public.dhe.ibm.com/software/data/cognos/documentation/docs/en/10.2.0/prfmdl_ug.pdf