Data Warehousing Techniques

and Standards
Presented by:
Davin Flatten
Systems Analyst
MassSAFE
University of Massachusetts

@
 Introduction
• The definition of a data warehouse
– What they are
– What they are not
• The star schema
– Dimensions
– Facts
• RDBMS requirements for warehousing
– ETL ( Extraction, Transformation, Loading)
– ROLAP ( Relational Online Application Processing )
• Keys to success

@
 What a data warehouse is not
• Not a OLTP ( Online Transactional
Processing ) system
• Not necessarily a physical place
• Not a single project with an end
• Not a single product or application

@
 What a data warehouse is
• Modeling technique specific to
analysis
• Bill Inmon’s four characteristics of a
data warehouse:
– Subject based design
– Integrated with your data
– Nonvolatile ‘picture’ of your states data
– Time variant view of your data

@
Where does the warehouse fit?

@
 Why have a data
warehouse?
• Supplements transportation data
systems
• Avoids conflicting resources
• Increases your organizations
understanding of your data
• Most importantly…

@
It completes the data life cycle.

@
 The Star Schema
• Key concept for data warehousing
• Modeling technique that simplifies joins
and tables
• Organizes data into a format that is
easy for business users to understand
• Allows application developers to
standardize ad-hoc queries

@
 Elements of a Star Schema
• Dimension Tables
– Easy to understand groupings of subject areas
– Can be hierarchal used to drill down
– Denormalized, decoded, and cleaned set of
descriptive data elements
• Fact Tables
– Contains foreign keys referencing dimension
records
– Contain either additive or semi-additive
measures for analysis

@
Sample Citation Schema

@
The Dimension Table
• Each record
contains a single
town (grain)
• The associated
information is
denormalized and
hierarchal
• All values are
decoded in plain
language

@
The Fact Table
• Each record
contains a single
violation (grain)
• Each dimension is
reference with a
foreign key
• Measures are
provided for fines
and violations

@
 The Star Query
• Regardless of
subject matter, the
same type of query
can be used
• Results can be
easily read and
used by analysts
• No complicated
outer-joins, sub-
selects, or other
complex SQL

@
Selecting a Relational Database
Management System (RDBMS)
• ETL (Extraction, Transformation, Loading)
– Ability to create stored procedures
– Job scheduler
– Logging and security tracking
• ROLAP
– Optimizer capable of performing star queries
– Table partitioning across time
– Bitmap index capabilities

@
 Why we use Oracle for our
warehouse
• Table partition pruning
• Star query optimizer hint
• Bitmap indexes
• PL/SQL stored procedures
• Transportable table spaces
• Query rewrite
• Materialized views
• Job scheduler

@
 Keys to a successful
Warehousing project
• Identified and involved warehouse
users
• Strong and committed leadership
• Diversified project team
• Established partnerships with all key
source data holders
• Incremental project plan the produces
fast results
• Correct design philosophy