Fiscal Year Crossover Time design vs. Generic Time design

Question

Hi all,
does any one know the differences between these two? And where are time design strategies documented.
Regards,
MK
Edited by: user4204276 on May 18, 2009 10:47 AM

Answer 1

MK,
I've never heard of "Genetic Time design". Google only returns your post -- see, you're famous. :)
Do you mean the difference between a Time dimension that contains years, months, weeks, days, etc. versus (crossover?) versus two or more dimensions (genetic?) to capture years and months? (Really, it could be any combination, decades and years, decades, years, and quarters, etc., etc.)
I've never seen any formal documentation that says one is better than the other, or really any documentation whatsoever re design considerations. In BSO databases, it's always been a question of balancing ease of calculation (usually) in a single Time dimension versus block size -- that often forces a split into two (or even more) dimensions to handle time. There, I guess that was a design consideration after all.
When Hyperion put together Planning, they split the Year and Period dimensions, for whatever that is worth.
In an ASO cube, block size is irrelevant/doesn't exist. Take a look at the pretty cool Intelligent Time (I think that's the right name) wizard that comes with an ASO cube -- that sucker can go right down to seconds with all kinds of built in intelligence. Those single Time dimensions are way bigger than any BSO cube could hope to hold.
From a reporting side year over year reports (qtr1 this year versus qtr1 last year) require dimension nesting -- this is generally not a big deal, but can be somewhat bothersome when many other dimensions are part of an asymmetric column selection.
Single time dimensions are way easier in calcs when doing something like a 13 month rolling average -- not having to write IF blocks to handle jumps across years is nice.
Lastly, unless you decide to go with a duplicate name database, you will have to come up with a naming convention that guarantees uniqueness in a single time dimension.
I'm sure others will (maybe they will) jump in with comments of their own -- if there's a section in the DBAG that explains which one to use I've forgotten it. <--This is entirely likely but I have to get up early for a flight so I'm not checking. :)
Regards,
Cameron Lackpour
P.S. One other thought -- if you must have a single time dimension with multiple years, quarters, months, and weeks, and you have to do it in a BSO cube, you're going to end up making Time sparse, with all kinds of calculation implications (slowness, chiefly). So there's another design consideration.