Database Reference
In-Depth Information
developers aren't so motivated by plugging gaps in a key performance
indicator (KPI) report. Data scientists, especially, are looking for new
insights, which is, by its very nature, a more experimental paradigm.
Model the Value
Once a valuable pattern has emerged, we have something we can apply a
schema to. However, don't just constrain your thinking when you see the
word model . In the world of analytics, that model may just be an algorithm
that has identified some interesting trends or shapes in the data.
A Hadoop developer may choose to harden the solution at this point and
also may “graduate” the results of the insights to a new Hadoop
environment for users to consume. This pattern is surprisingly common.
Don't assume that there is only one Hadoop cluster to support one
production environment.
In Figure 10.2 , you can see an example of what the Hadoop environment
mightlooklikeandhowthedatacouldgraduatebetweenenvironments.The
far-left bar represents the Bronze environment. This is the raw data and is
an environment typically used by the “data scientist” mining for value. As
patterns emerge and value is established, the results might find themselves
moving into the Silver environment for “power users” or “information
workers” to work with. There are usually more power users than there are
data scientists. Finally, the data may go through one last graduation phase
from Silver to Gold. This might be the environment exposed to the broadest
spectrum of the business (that is, the “consumers” of the data). If there
are dashboards to be built and so forth, it is likely that Gold is the target
environment to perform these tasks.
 
Search WWH ::




Custom Search