The person happened to be the BI/Analytics Application owner at a $15B global manufacturer.  They had (4) primary data warehouses – Finance, Sales, Customer Service, and Tech Support.

For each warehouse, they had (3) “stacks” or silo’s – Development, Test, and Production.

Their stated problems were:

- Redundancy of data across all four warehouses and all 12 stacks

- Redundancy of infrastructure for the same

- Inconsistent architectures (HW and SW)

- Aging infrastructure

Their problems meant:

- People scale issues – 50% of the time of a BI analyst spent doing data collection and reconciliation.  Worse, it takes 6-12 months to train an analyst on the tools for an average retention rate of 18 months!

Their goals were:

- Consolidate to a single master data mart

- Consolidate to a single master physical infrastructure stack

- Standardize on a SW architecture

- Develop a “virtual” data layer for reporting and analytics

What they have found:

They spent almost 18 months re-architecting the back end to consolidate the physical data repositories.  Now instead of scaling by adding new databases, they make the ones they have bigger.

By developing a common virtual front end for reporting and analytics on top a common data back end, they were able to standardize on tools and process – which they think cut analyst training time down to 3 months or less.

By physically eliminating excessive duplication of data, they spend 80% less time reconciling.

What I found MOST interesting:

1.  The way they collapsed the Dev/Test/Production stacks was enabled by HARDWARE functionality.  They currently only do one load per day, but by putting all the data on a common platform (V-Max in this case) and using BCV’s, they can now do (4) loads daily – without screwing up production.  Almost in passing he mentioned that the Development and Test people became twice as productive overnight.  We forget how little things we take for granted coming from the infrastructure side of the world can mean to our application pals, who tend to think all infrastructure is evil.  Since this hardware function isn’t exactly new, I’m guessing we stopped trying to tell our app pals how they could leverage this and live a longer, healthier life!

2.  Using the Oracle Linux stack they went from a bunch of huge Sun Unix machines to 7 Dell servers running RAC.

3, Total Project Payback – 16 months.  Total HW/SW payback – 8 months.

4.  He said “consolidation” a hundred times – and not once did he mean infrastructure.  He was talking about consolidating application and data instances.  I hear consolidation and I’m thinking boxes.  He didn’t care about boxes.  This is another example of the importance of finding understanding in the motivation of different groups within IT.  The guy selling the Dell servers most assuredly heard consolidation to mean “get rid of the Sun boxes and buy some of mine”.  The Oracle guy heard, “get rid of all those disparate tools and standardize on my consolidated suite”.  The IT infrastructure guy heard “get rid of the old gear and consolidate onto new infrastructure.”  What the App owner said, however, was “let’s consolidate data marts and standardize our architecture.”

It’s all in how you say it, I guess.