Databases Reference
In-Depth Information
The warehouse server, or
enterprise data warehouse
, is a multisubject historical infor‐
mation store usually supporting multiple departments and often serving as the corpo‐
rate database of record. When an ODS is established, the warehouse server often extracts
data from the ODS. When an ODS isn't present, data for the warehouse is directly
extracted and transformed from operational sources. External data may also feed the
warehouse server.
As noted previously, platform consolidation is popular within these tiers today. The
enterprise data warehouse can be the point of consolidation for the ODS and multiple
data marts. Although different logical models remain, they are consolidated to a single
platform and database.
OLTP Systems and Business Intelligence
True real-time data resides in the OLTP systems. Organizations can provide reporting
from transaction processing systems side by side in portals or dashboards with infor‐
mation from data warehouse systems. A key to providing meaningful dashboards is to
provide high-quality data with consistent meaning. The quality of data in OLTP systems
is directly related to controlling data input to eliminate duplicate or error-prone entries.
Consistent meaning can be resolved using master data management (MDM) solutions.
MDM solutions consist of data hubs that serve as a common reference point for data
supporting key business measurements such as customers, products, or finance. Oracle
offers a number of data hubs for these and other business areas to enable building out
of such an infrastructure.
Projects that include data from data warehouses, OLTP systems, Big Data sources, and
MDM solutions are called data integration projects. Most business intelligence deploy‐
ments, at the time of publication of this edition, use just the data warehouse infrastruc‐
ture as the primary source of historic data for business intelligence. The extraction,
transformation, and loading (ETL) techniques applied to the data warehouse are de‐
signed to resolve differences in common data elements, to cleanse the data, and to
provide a historical database of record.
Big Data and the Data Warehouse
Organizations are considering extending the business intelligence topology as they in‐
troduce Big Data platforms. These platforms are commonly defined as those that run a
distribution of Apache Hadoop, an open source software framework ideal for analyzing
unstructured or semi-structured data. At the base of any Hadoop cluster deployment is
the Hadoop Distributed File System (HDFS) for storing the data and MapReduce for
determining data of value. Other Hadoop software components that are often part of a
deployment include:
