Operational Data Store
Need
to create an operational data store for an information management
project and want an overview of key design considerations?
What is a
data warehouse?
A data warehouse is the foundation for information management and
involves:
- Extracting data from operational systems;
- Moving it into data warehouse structures;
- Reorganizing and structuring the data for
analysis purposes; and
- Moving it into reporting structures called data
marts.
What is an operational
data store (ODS)?
An operational data store is a collection of operational data which is
used to support operational monitoring and management. It involves:
- Extracting data from operational systems;
- Moving it into ODS structures; and
- Reorganizing and structuring the data for
analysis purposes
How is an ODS different
from a data warehouse?
A data warehouse is intended to support strategic planning and business
intelligence decision support. It should contain:
- Integrated subject oriented data, e.g sales data;
- Static data, e.g. data that is moved into data
warehouses should not change after it is stored in the data warehouse
environment;
- Historical data, e.g. data warehouses will usually
contain several years worth of historical data; and
- Aggregated,
or summarized data e.g. as data becomes "older", it is summarized to
reduce data storage requirements and to improve analysis performance.
An operational data store is intended to support operational management
and monitoring and should contain:
- Integrated subject oriented data (similar to data
warehouses) e.g sales data;
- Volatile data, e.g. data that is moved into an ODS
will probably change frequently;
- Current
data, e.g. an ODS will usually contain several weeks or even months
worth of data instead of large volumes of historical data; and
- Detailed data e.g. as data becomes "older", it
is
summarized to reduce data storage requirements and to improve analysis
performance.
ODS data is refreshed frequently to provide a "snap-shot" view of the
on-line transaction processing (OLTP) systems and legacy systems.
Who uses an ODS?
An ODS is frequently used by customer support staff
to access integrated data to respond to customer inquiries.
What are ODS design
considerations?
An ODS should be based on an normalized relational database design,
similar to a data warehouse design, usually with no data aggregation
structures.
What are ODS data load
considerations?
Typical methods of loading an ODS include:
- Extract,
transform and load (ETL) software e.g. Informatica, Ab Initio,
Ascential DataStage is used to collect source data and populate ODS
structures;
- Materialized views, which create a snap-shot of
source tables and can load custom designed views for reporting
purposes; and
- Data replication software e.g. Oracle Data Guard can
be used to populate ODS structures that are identical to the source
system or they can even be used to create, and populate logical, or
more normalized, structures.
Requirements analysis
considerations?
Do not build an ODS in hopes that some-one will want to use it!
ODS requirements should be based on approved data requirements and a
logical data model.
Summary…
A data warehouse contains a large volume of historical, static, data to
support business intelligence decision making.
An operational data store contains a smaller volume of current,
volatile data, to support operational management and monitoring.
|
