Database Reference
In-Depth Information
stability, flexibility, and performance in GPFS-FPO, as well as all of the utili-
ties that you're used to. GPFS-FPO also provides hierarchical storage man-
agement (HSM) capabilities, whereby it can manage and use disk drives with
different retrieval speeds efficiently. This enables you to manage multitem-
perature data, keeping “hot” data on your best-performing hardware. HDFS
doesn't have this ability.
GPFS-FPO is such a game changer that it won the prestigious Supercom-
puting Storage Challenge award in 2010 for being the “most innovative stor-
age solution” submitted to this competition.
NOTE
GPFS-FPO is currently available only for beta testing.
Wrapping It Up
We said it earlier in this topic, but it's really important to understand that Big
Data does not solely equal Hadoop. Hadoop is but one of multiple data pro-
cessing engines you will need to address today's challenges (and the ones
you don't yet know about yet). For this reason, IBM has been long committed
to Hadoop, with contributions to the Apache project and continual engineer-
ing around the ecosystem. With its long history of enterprise-grade infra-
structure and optimization, IBM has taken this experience and applied it to
Hadoop through BigInsights. BigInsights includes open source Hadoop and
adds some operational excellence features such as Big Data optimized com-
pression, workload management, scheduling capabilities, and even an app
development and deployment ecosystem.
While operational excellence delivers economics of scale and infrastruc-
ture trust, the true potential of Hadoop lies in the analytic capabilities—more
accurately, the consumability of the analytic capabilities. For this reason,
BigInsights includes toolkits, accelerators, and tools such as BigSheets, which
seek to democratize the use of Big Data so it isn't bottled up in deeply skilled
tech departments that require an army of Java programmers to monetize the
data. In addition, BigInsights provides an end-to-end text extraction frame-
work. Quite simply, Hadoop doesn't provide anything here, so IBM builds a
high-value stack that treats Hadoop as a first-class citizen and integrates it
across the IBM Big Data platform.
