Implementation of Big Data Analytics - Big Data Analytics: Disruptive Technologies for Changing the Game - page 63

The hybrid approach promoted actively by IBM's Information Agenda team
places the AAP architecture on top of existing BI infrastructure. All the Big Data
lows through AAP, while conventional sources continue to provide data to the
data warehouse. We establish a couple of integration points to bring data from
the warehouse into the analytics engine, which would be viewed by the data
warehouse as a data mart. A sample of the AAP data would be directed back
to the data warehouse, while most of the data would be stored using a Hadoop
storage platform for discovery. The hybrid architecture provides the best of
both worlds; it enables the current BI environment to function as before while
siphoning the data to the AAP architecture for low-latency analytics. Depending
on the transition success and the ability to evolve skills, the hybrid approach
provides a valuable transition to full conversion.
Both the revolutionary and the hybrid architectures signiicantly challenge
the data governance function. The next section describes the new set of issues
and how to handle them.
6.2 Big Data Governance
Three broad categories of questions are emerging in the area of Big Data
governance:
• Single view of the customer —We now have access to more complete data
on how customers use their products for their communications, content,
and commerce needs. How do we merge this newly acquired data with
everything else we have been collecting to create a more comprehensive
understanding of the customer?
• Big Data veracity —Customer data comes from a variety of “biased”
samples with different levels of data quality. How do we homogenize this
data, so that it can be used with conidence?
• Information lifecycle management —This is a lot more data than we have
ever encountered before. Our current analytics systems are not capable
of ingesting, storing, and analyzing these volumes at the required
velocities. How do we store, analyze, and use this data in real-time
or near real-time?
We will use this chapter to elaborate on these questions and will provide partial
answers as they are known today.
Integrating Big Data with MDM
During the 1980s and 1990s, we created a series of departmental applications
based on business cases associated with workforce automation. The result was