We are using Hyperion Planning 11.1.2. Our business model is based on 20 Dimensions. We have a "SPARSE" dimension where we have huge number of members e.g. approx 12,000 dimension members. Now, i want to confirm that, if we reduce these member by re-visiting our business model then it will be reduced from 12,000 to 300 or 500 members.
How it would impact on Application's Overall performance???
I am sure less is better that more if you don't need more.
less sparse members mean less potential number of blocks. It can improve some forms and scripts performance.
Are you sure you want to deploy your app with 20 dimensions. This is not a best practice at all. May be you can review the model and try to reduce the total dimensions based on some OLAP and Essbase advantages like drill down, attribute dimensions, UDA's, alternate heirarchies.
Reducing dense members is I think big deal compared to sparse as it deals with block size.
Obviously, member reduction on Dense side effect more on performance. But this is our live environment and we are facing performance issues frequently. Currently, Hyperion architecture is 32bit and we are also thinking it to upgrade on 64bit. So, we want to upgrade to 64bit by reducing members as much as we can.
Any more advice???
Ask yourself the following questions:
1. What is the application used for? Do you need users entering or modifying data? Do you have any calculations running outside of the outline formulas.
---- All these will let you decide whether you can use an ASO, which will allow you to keep all the members and all the dimensions.
---- If you've calculations in your cube that can be done outside before the dataload, the ASO application is still feasible
2. If ASO is not an option, then reducing the number of members will reduce your database size considerably (especially because you've 20 dimensions) ... It will also help your calculation time, maintenance time (defragmentation, data cleanup etc). However, unless you've rest of the dimensions with very few members, 20 dimensions is not a good idea. You can also look into your block size and see if you can reduce the number of sparse dimensions that you've by making another dimension dense.
When I design anything, I think of ASO first and only if I can't have an ASO, I go to BSO.
Did you mean 64-bit s/w or server upgrade? If it is the hard ware you can surely expect better performance and it will let you assign more java through put on these processes. I heard complaints about the performance degrade after upgrading servers as well. I don't mean to scare you :) , but there is no general answer for these questions. In theory it should work very well and improve the performance statistics.