Database Reference
In-Depth Information
you use in your modeling?” and “How long does it take to iterate and refresh
over these models?” Current modeling methodologies are held back by sys-
tem constraints, which dictate the architecture and preclude pursuing the
path to optimal results. It's not that companies don't recognize that there are
a lot of potential insights available from undiscovered data, but their current
systems don't always support them. Doubling, tripling, or quadrupling the
size of an analytic warehouse isn't going to work, because they're usually
already bursting at the seams or fully utilized. Capital constraints and regu-
latory requirements are forcing adoption of interesting new approaches in
this space, and we thought we'd share some of them with you in this section.
Consider a multinational client that needed to move from line-of-
business-organized stovepipes to individual-level risk management. Their
current data is a month-old snapshot of their credit exposure; useful data for
sure, but slow and expensive to bring through their current manual process.
We proposed a new architecture that includes BigInsights, IBM Information
Server, and Netezza (the name of the IBM PureData Systems for Analytics at
the time). We used their existing IBM Information Server Data Stage plat-
form to transform the raw data, and loaded the enriched data into HBase
(a column-oriented data store that's included in the BigInsights product).
HBase enables you to persist and present data as a key/value pair. Client
credit exposure was written out as a time series that was expanded as new
products were offered. Using this approach, BigInsights was able to maintain
a current representation of the client's credit positions in an environment
that's both elastic in sizing and performance, and significantly less expen-
sive. The Netezza analytics environment can request data from BigInsights,
whenever needed, to enrich modeling without having to be concerned with
how the client evolved the process of acquiring and storing their data.
Wrapping It Up
We can't possibly do justice to this topic in such a short amount of space. The
bottom line, however, is that having more data can't just give you deeper
analytical insights, but actually accelerate the analytics process as a whole—
you just need to have a platform that helps you to contain the side effects of
more volume, variety, veracity, and velocity—more data. The IBM Big Data
