Forum Discussion

camagruder's avatar
camagruder
New Contributor III
3 years ago

Does the "In Use" tag impact Data Unit Size?

We are looking for ways to reduce the data unit size in one of our cubes in order to increase reporting performance without having to do a redesign.  We currently have +20k UD1's and +10k accounts in...
Metadata
  • PFugereCSO's avatar
    PFugereCSO
    3 years ago

    Hi Christy,  

    As for zeros source, I would be just taking an educated guess.  But you are right to want to explore that.  Some people find the No-Data for zero settings create a situation where they feel they have to load zeros. Ive been guilty of this too, but it can be a problem.  A zero is not a NoData value.

     

    As for not disrupting the end users - It is always best practice to have at least two cubes - rarely would you only have one.  A detail and summary cube.  This allows to make an update like what you need here without impacting the end users.  If you only have one cube, then your options are getting limited.  With two cubes its easier to update the metadata for the parent cube.  IF you are working with RCS, they can help you.  I am happy to help too off this thread if you want to brainstorm design ideas.  

camagruder's avatar
camagruder
New Contributor III

Hi Peter - thanks for the reply!

We have been reading your performance guides and I really appreciate all the detail and information that you and Tom have provided.  The large client that was mentioned in the performance white paper seemed so familiar ğŸ˜€

We have been combing through the system diagnostics dashboard looking for opportunities to improve performance.  It appears most of the zeroes in the forecasts are where data from actuals were seeded into the forecast since actuals are hitting many more accounts than the forecast.  So if you look at a February forecast, January actuals are causing a lot of zeros in the forecast months (Feb-Dec) since there is nothing forecast in those months.  Once we get to a December forecast, we have much less zeroes (less than 10% which I believe was your recommendation).  Should we be looking at our seeding logic or are these type of zeroes unavoidable?

I am not exactly sure how we would explore better use of dimension extensibility without disrupting users.  I might need to explore that further.  Do you have any suggestions?

We are starting to play around with hybrid scenarios this week so I am also hopeful that might be a possibility that we can start moving some of our users toward for reporting.  

Thanks again,

Christy

 

 

 

 

PFugereCSO's avatar
PFugereCSO
Contributor
3 years ago

Hi Christy,  

As for zeros source, I would be just taking an educated guess.  But you are right to want to explore that.  Some people find the No-Data for zero settings create a situation where they feel they have to load zeros. Ive been guilty of this too, but it can be a problem.  A zero is not a NoData value.

 

As for not disrupting the end users - It is always best practice to have at least two cubes - rarely would you only have one.  A detail and summary cube.  This allows to make an update like what you need here without impacting the end users.  If you only have one cube, then your options are getting limited.  With two cubes its easier to update the metadata for the parent cube.  IF you are working with RCS, they can help you.  I am happy to help too off this thread if you want to brainstorm design ideas.