How to Use A Fast, Data-Driven Metadata Load for a "Big Data" Data Warehouse Appliance Project

作者：Anonymous

来源：Database Trends and Applications, 2011, 25(1): 15.

Recently, Melissa Data had a client that required migrating and re-hosting a data warehouse from the Oracle environment to a Data Warehouse Appliance. The application had hundreds of tables and almost 2,900 Source-toTarget Mappings. This would have taken thousands of hours of manual coding. For this client, we implemented a reusable data-driven architecture that relies on a Metadata Mart, to generate the load processes. The end result was a greatly reduced TCO (Total Cost of Ownership) for generating the code required vs. manual coding. Gartner’s Michael Blechar states, "Best practices include scoping metadata management into smaller areas where governance can be more easily applied at a more abstracted level of detail in repositories which I call ’metadata marts’" in his recent blog post, Down With the Uber-Repository, Long Live Metadata!1 The objective was to enable the client’s analytical and reporting tools to access their data via the Data Warehouse Appliance platform quickly and with minimal coding effort. We loaded the client’s data into the Data Warehouse Appliance platform as fast and inexpensively as possible while still providing a reusable "production level" set of code for implementation.