This is peng Wenhua’s 182nd original article \
In fact, there are a lot of modeling articles, but they still stay in what star type, snowflake type these relatively superficial content level. \
In fact, modeling is a very demanding technical task. And this work is not to say that the company cattle, technical cattle can be done, this matter directly determines the success of the project.
In February 2021, IBM lost hundreds of millions of dollars to customers after a data warehouse project failed. One reason was that the people IBM asked didn’t follow Teradata’s Financial Services Logical Data Model (FSLDM) to design the ENTERPRISE data warehouse EDW.
As a melon eater, I like to watch. Also, the modeling can tell you a lot about it.
What is this project?
The reason is that Direct Line, a British insurance company, launched a project “Best for Customer” in 2014. Like all insurance companies, the core of this project is to serve customers. \
They wanted to put all their customer data in EDW, an enterprise data warehouse, and then build a new platform on top of it, and all the insurance business would run on top of it. Through data, business processes can be smoother and customer value can be maximized.
Look, isn’t that a good idea? There’s nothing wrong with it!
The architecture of the whole project was already designed. Teradata Database14 is used for data warehousing, and Informatica’s products for data migration and ETL are among the world’s best. The data model is Teradata’s Financial Services Logical Data Model, FSLDM.
Some of you are unfamiliar with these two products. ETL tool just said, Teradata is fun. To tell you the truth, Teradata was once the biggest data adata company in the world. I didn’t call it that. The client did. Teradata has long been an industry leader in building data models.
So there’s a problem?
In theory, familiar business, powerful technology and the formidable combination of IBM, Teradata and Informatica should give it time. Although there are some new system construction and switching between old and new systems, there are perfect solutions and there shouldn’t be any problems.
But it was precisely the project that couldn’t go wrong that ended up costing IBM money.
At the heart of the dispute is Direct Line’s claim that IBM failed to follow Teradata’s Financial Services Logical Data Model (FSLDM) design. While Teradata already has the standard model, it has to repeatedly build existing entities. \
Quote: “Extending the model by copying and pasting in a disjointed and unsupported manner results in breaking the design integration layer and making EDW difficult to populate, maintain, and understand.” This comment is really a breakdown.
It was obvious there was something irreconcilable between IBM and Teradata. My guess was that IBM was going to do the project, but the relationship didn’t work out, Teradata never gave him any data or support, and IBM went it alone. Working on his own, without Teradata backing him, he had to build from his own experience, which eventually crumbled.
In the end, IBM handed over the entire code in 2016, leaving Teradata to take over and start over. That’s what you’re calling it.
We’re not done yet. The owner is suing IBM! The lawsuit went on for years. It’s a two-way street anyway, so just say you’re okay. This case was not awarded until February this year. I think it cost more than 300 million yuan.
I don’t care about IBM at all. What I want to share with you is Teradata’s Financial Services Logical Data Model, FSLDM. That’s a tough one. There’s no motivation.
If there are more than 30 “looking at” in this article, I’ll write a single chapter that dissects Teradata’s standard data warehousing model, FSLDM, and shows how the industry’s most classic data warehousing model is built.
Enjoy better with the following articles
Dry goods | breath finished data warehouse modeling method
Dry goods | how to build a data warehouse
Data warehouse construction complete information package \
【 real time 】 Build a real-time data warehouse hand by hand
[dry goods] how many storehouse to divide after all?
Why do data warehouses have AN ODS layer? \
