Scribd
Upload
247 views16 pages

1030 Stephen Brobst Semantic Data Modeling

Semantic data modeling is a logical data modeling technique. A semantic data model captures the business view of information for a specific knowledge worker community or analytic application. Logical data models capture the business relationships in the enterprise information. Physical data models capture implementation design of tables in the data warehouse.

Uploaded by

haiderabbaskhattak
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
247 views16 pages

1030 Stephen Brobst Semantic Data Modeling

Semantic data modeling is a logical data modeling technique. A semantic data model captures the business view of information for a specific knowledge worker community or analytic application. Logical data models capture the business relationships in the enterprise information. Physical data models capture implementation design of tables in the data warehouse.

Uploaded by

haiderabbaskhattak
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 16

Semantic Data Modeling:

The Key to Re-usable Data


Stephen Brobst
Chief Technology Officer
Teradata Corporation
[email protected]
617-422-0800

Enterprise Information Management


Data Modeling

Not just a collection


of subjects...

...but also their


relationships
Single, Integrated System

Party

Account

Product

Activity

Party

Dont model subjects


individually!
2

Product
Account
Activity

Model your entire


business!

Copyright 2013. Stephen Brobst. Do not duplicate without written permission

Functional Views

Sales

Marketing

Finance

Rates/
Regulatory

Customer
Service

Risk

Promotions
Demographics

Pricing
General Ledger

Contracts

HR

Engineering

Safety

Products

Production

Works OK for OLTP, but causes


data chaos for BI applications.
3

Copyright 2013. Stephen Brobst. Do not duplicate without written permission

Business Intelligence Requires Data Integration

Product Data
Customer Data
Account Data
Transaction Data
G/L Data
Market Data
External Data

Copyright 2013. Stephen Brobst. Do not duplicate without written permission

Data Modeling Techniques


Key observation: Practitioners in the data warehousing
industry frequently confuse construction of the semantic data
model, logical data model, and physical data model.
A semantic data model (SDM) captures the business

view of information for a specific knowledge worker


community or analytic application.

A logical data model (LDM) captures the business

relationships in the enterprise information independent


of a specific analytic application or departmental view.

A physical data model (PDM) captures the

implementation design of tables in the data warehouse.

Copyright 2005, Stephen


A. Brobst.
All rights
reserved.Brobst. Do not duplicate without written permission
Copyright
2013.
Stephen

Data Model Deployment


Conceptual Data Model

Enterprise Logical Data Model(3NF)


Subject Area
B

Subject Area B
xxxxxxx
xxxxxxxxxx
xxxxx
xxxxxxx

xxxxxxx

xxxxxxx
xxxxxxxxxx
xxxxx
xxxxxxx
xxxxxxx
xxxxxxxxxx
xxxxx
xxxxxxx

xxxxx
xx

Subject Area
A

Subject Area A

xxxxxxx
xxxxxxxxxx
xxxxx
xxxxxxx

Subject Area
C

xxxxxxx
xxxxxxxxxx
xxxxx
xxxxxxx

Design Meta
Data
Enterprise Data Standards

Physical Model Realization


Single Physical Data Model

Project A

Project B

Project C

Semantic Model Views


6

Copyright 2013. Stephen Brobst. Do not duplicate without written permission

Semantic Data Modeling


Semantic data modeling is a logical data

modeling technique; the semantic view of


information does not necessarily need to be
physicalized in the database.

There may be a different semantic data model for

each department/applications that uses the data


warehouse.

Dimensional modeling is a common technique for

constructing the semantic data model for an


analytic application, but is not the only viable
approach.

Copyright 2005, Stephen


A. Brobst.
All rights
reserved.Brobst. Do not duplicate without written permission
Copyright
2013.
Stephen

Different Semantic Model Designs are Appropriate


for Different Types of Knowledge Workers
Dimensional

Relational

ADS

Application

Physical Data Extensions


Index choices &
selective table
denormalizations

Normalized Generic Structures

Copyright 2013. Stephen Brobst. Do not duplicate without written permission

Physical Data Model


Physical data model represents the tables constructed in
the database.
Recommendations:
Use the (3NF) LDM as the starting point for the PDM

with selective denormalization when appropriate for


(primarily) performance reasons.

Overlay (dimensional) SDM on top of PDM using

views and/or semantic metadata in your BI tool.

Design LDM first, then use application-specific

business requirements to derive the SDMs and


performance considerations to map into the PDM.

Copyright 2005, Stephen


A. Brobst.
All rights
reserved.Brobst. Do not duplicate without written permission
Copyright
2013.
Stephen

Semantic Models Should be BI Tool Agnostic

MicroStrategy
Teradata OLAP Connector

Tableau

10

Tier 3

Access

Tier 2

Integrated

Tier 1

Acquisition

Copyright 2013. Stephen Brobst. Do not duplicate without written permission

What is a Semantic Modeling


Building Block (SMBB) Portfolio?
A collection of data modeling assets that help make database
design and development faster and easier for the access
layer:
> Access layer provides path for data from the integrated data model
to end user consumption.
> When this layer not well-designed, it can impact speed, security,
and simplicity in developing and delivering reports, BI applications.

Re-usable building blocks provide flexibility and consistency


to the development process:
> SMBBs include pre-built semantic models.

Focuses on a specific analytic need


in a specific industry:
> For example, Communications Mobile
Revenue Analytics.

SMBBs are to the semantic layer as


iLDMs are to the integrated layer of
a data warehouse implementation.
11

Copyright 2013. Stephen Brobst. Do not duplicate without written permission

Dimension Building Blocks Support a


Range of Analytical Needs
Dimensional Model

Dimension Building Blocks

Fixed, Flattened Hierarchy

12

Variable Depth Hierarchy


(Recursive)

Fixed, Normalized Hierarchy

Copyright 2013. Stephen Brobst. Do not duplicate without written permission

What are SMBBs?


How are they related to an LDM?
Building from the Foundation for your
Data Warehouse:
An LDM is like a blueprint for a house
that you are building. It serves as the
foundation for your integrated data
warehouse.
The SMBBs are like room designs that
meet specific homeowner needs.
Different rooms need different designs
based on their purpose. Similarly, for
each new business application, new
semantic models are needed.
SMBBs provide different designs
(building blocks) for the modeler to
choose from in building the semantic
models.
These flexible, reusable building blocks
can be used for other analytic needs.
13

Copyright 2013. Stephen Brobst. Do not duplicate without written permission

Relationships between the


Three Types of Data Models
Q: Where does it all start?
A: Business requirements drive the process!
Logical Data Model

Support
data
re-use

14

The Logical Model is


used to drive
generalization and
support source data
leverage and reuse.

Semantic Data Models

Physical Data Model

Data
access
patterns

The Physical Model


provides core
support for data
integration within
the information
architecture.

Copyright 2013. Stephen Brobst. Do not duplicate without written permission

The Semantic Model


captures data
access patterns that
must be supported
by the core physical
model.

Semantic Layer Benefits


Efficient table joins can be encouraged
inside the SDM views.
Views are low maintenance objects.
Views do not consume database space.
Join indexes (JIs) and aggregate join
indexes (AJIs) can be created based on the
access paths embedded in the SDMs.
PDM is not compromised with new
application requirements.
Protection of code assets.

15

Copyright 2013. Stephen Brobst. Do not duplicate without written permission

Conclusions
Critical to distinguish between logical data modeling,

semantic data modeling, and physical data modeling.

Separate the implementation of the semantic model

from the physical data model (PDM) deployment for


maximum flexibility.

Selective use of PDM extensions to optimize

performance.

Either ANSI standard views of the semantic metadata

within your BI tool of choice can be used for creating


a semantic data layer without sacrificing flexibility of
the PDM.

16

Copyright 2005, Stephen


A. Brobst.
All rights
reserved.Brobst. Do not duplicate without written permission
Copyright
2013.
Stephen

You might also like