TM1 to Dynamic Cube

[froodyo]

Hello Everyone,

I am struggling with extremely slow TM1 reporting cubes, I have upgraded hardware, bumped up memory, and started re-writing reports to be as efficient as possible, i.e. getting rid of any detail filters and using dimensional functions wherever I can. But, I am really curious about two things:

1. Is it possible to create a dynamic cube from TM1 dimensions and/or cubes?
2. Would I see a performance boost with using dynamic cubes over just hitting a TM1 directly with Cognos Anlaytics?

Sorry about the question being so broad, but I am just at a loss with how slow these cubes are to report against and I am not sure how to calculate the size of them to give specific details on the cube itself.

Thank you,
Grady

[MFGF]

Hello Everyone,

I am struggling with extremely slow TM1 reporting cubes, I have upgraded hardware, bumped up memory, and started re-writing reports to be as efficient as possible, i.e. getting rid of any detail filters and using dimensional functions wherever I can. But, I am really curious about two things:

1. Is it possible to create a dynamic cube from TM1 dimensions and/or cubes?
2. Would I see a performance boost with using dynamic cubes over just hitting a TM1 directly with Cognos Anlaytics?

Sorry about the question being so broad, but I am just at a loss with how slow these cubes are to report against and I am not sure how to calculate the size of them to give specific details on the cube itself.

Thank you,
Grady

Hi,

The way TM1 cubes work is that they read all the data and load it into memory, and as a result of this memory-resident architecture they can be very responsive with reasonable sizes of data, only starting to slow down when the data volumes get large. However even small cubes can perform poorly if you don't optimize them - they will by default try to calculate values for each intersection, even if they are empty. I remember some of my TM1 colleagues a few years ago telling me about skipchecks and feeders, and how cubes should never be overfed - but bear in mind I'm not a TM1 expert so things might have moved on since then without me realizing. As a first step I'd be looking at whether you can optimize the TM1 cubes - you'll probably need to get some help from a TM1 expert with this.

Dynamic Cubes are an entirely different technology than TM1. They are firstly designed to work over the presentation layer tables in a data warehouse - ie you will need to have a data warehouse with star or snowflake schemas. There is some in-memory caching technology to optimize performance, and also they can use in-database aggregate tables or views to speed up performance too. Because not all the data is memory-resident, they can scale to much larger sizes than TM1 cubes and still give really good performance - albeit they need to be optimized to perform well, with appropriate in-memory aggregates and in-database aggregates used.

Although both technologies share the same dimensional principles, they do not share a common metadata approach. What this means is that you can't take the model for a TM1 cube and build a Dynamic Cube from it (or vice versa). You would need to manually re-create the dimensions and measures from your TM1 cubes in Cube Designer if you want Dynamic Cubes of the same structure.

Also bear in mind that the Member Unique Names (MUNs) will be different between a TM1 cube and a Dynamic Cube, so some of your reports might break if you try to swap one out for the other.

Cheers!

MF.

[froodyo]

Thanks MFGF, that helps clear that up and points me back in the right direction.

Thanks!