Wednesday, February 4, 2009

Is your architecture leading to Stovepipe data mart?

Question arose while we were discussing BI. I asked myself, whats the hell is "Stovepipe" data mart, I was aware of the word but picture behind it, did not come to my mind immediately. So, brushed up.... Stovepipe data marts are nothing but data marts that prevent integration of themselves for making a data warehouse because of the unavailability of conformed dimensions. In most cases, data marts (dimensions in data marts) are designed with their own rules, policies and mainly own structures. This results incompatible dimensions in different data marts though they represent same entity. It will be a mess, will be awkward situation when it comes for making a data warehouse, combining all data marts. Though data marts have many advantages, cost-wise, time-wise, etc.. this is the major disadvantage of them. Are you leading to a Stovepipe data mart?

1 comment:

Gogula G. Aryalingam said...

When designing dimensions for a data mart, I feel that we have to take into consideration the entire organization instead of only that department or section of the organization for which the data mart is being built.

The dimensions should contain attributes that are being used in the entire organization and not just the attributes of the current scope. This of course would take a little more time and resources, but hey in the future it would ease the complexity in integrating the marts into a warehouse.

Then again its easier said than done. Most development houses and PMs push for shorter development times and faster delivery schedules, which most of the time leads to a stovepipe.