LA, L0, L1, L2 Data Warehouse Architecture

Hi,

Just need your opinion about the following data warehouse architecture for a bank that was implemented by a Vendor (using the best practices in DW development and lifecycle)

There is nobody in-house that has any data warehousing knowledge, until I joined (with 11 years of experience). After over 1 year of development, and huge amount of money spent, the vendor delivered the following DW environments.

Our current state:
LA - landing area
L0 - stage area
L1 - consolidated data, but not with fact or dimension tables
L2 - dimensional model for reporting

So, here are my issues with what they call the best practices in DW development.

1. L0, L1 and L2...Did anyone hear about these types of labels before? Is it a well known standard? Is it a european way of separating dw layers ? I know it means: stage (L0), integration (L1), presentation (L2), but don't really understand the labelling. What is the purpose of LA (landing area) if we have have L0?

2. According to our vendor, the L1 stage does not really contain dimensions or facts, even thou they have a snowflake schema like design. Party (dim), Product (dim), Service (dim), Country (dim), Address, Currency (dim), Transactions (fact), etc...but they insist this is not the DW layer that should have dimensions and facts, even though they are applying SCD type 2 to a lot of tables...I am not sure what it is anymore...kind of confusing...Did anyone have any experience with this type of L1 integrated layer.

3, L2 practically does not exist at this point. There are only 5 tables not in a dimensional model format at all. Basically, pre-aggregated tables for CRM purposes.



Therefore, majority of our reporting is done from L0!!!!! Yes, you are reading it right....reporting out of L0 ladies and gentleman...Because it costs more money to bring data to L1, not to mention L2.

From my experience, dim_Date is one of the first dimension tables created, this table is not there at all....just frustrated with this Vendor's approach and development practices that they call the best industry practice.

I am just looking for your comments in terms of the LA to L2 architecture and if the vendor has the right approach to this DW development.

I feel that the vendor created a monopoly and provides extreme estimates for new development and maintenance that the bank is ulimately paying.


Your thoughts/opinion are appreciated.

Regards,

RKJR

Can you get a refund? Unfortunately, I am hearing more and more of these types of stories. L2 is the whole point of the data warehouse. If that's not complete then all the other pre warehouse reporting staging is just that. Most of my work used to be building data warehouses. My work now is rebuilding "best practice" data warehouses.

'Best Practice' is a term open to interpretation. They are implementing an Inmon architecture, which tends to have longer development cycles and is more costly to implement.

As far as landing areas go, it usually refers to a location where external applications leave the data for processing by the DW load. This could be a disk somewhere to collect files, or it could be database tables or both. It is a common approach in an ELT environment when most transformations take place on the database. It is less common using traditional ETL.

thank you for your feedback..very much appreciated.

You are being ripped off.

Is this a Big Consulting company?

Yes, quite a big consulting vendor....they usually win tenders over IBM or Oracle in Europe....

That's their M.O.
- Bring in a lot of staff (mostly offshore or underqualified graduates) with one or two people who know what they are doing
- Extend the project timelines
- Reassure Senior Mgmt that they are getting "Best Practices".

I've worked in companies where these Big Consulting companies have wasted millions of dollars on delivering substandard solutions. How about 1 Billion dollars to deliver a solution that was so poor that it was scrapped and replace with an Access database?

I did a Proof of Concept for a bank with one of my staff members and in a week we had built ETL, a datamart, reports, dashboards and cubes using the Microsoft BI stack. The Bank had tried to deliver something with Oracle and B.O. for three months and didn't have anything to show for it, When we showed my parent company at the time (a Big Consulting company), the response was "Why would we want to deliver something in a week when we could take 3 months or more?"

really appreaciate your feedbacks...