Database Reference
In-Depth Information
does not restrict the developer to a single hop. Complex multihop flows
are perfectly possible with Flume. This includes creating fan-out and fan-in
flows; failover routes for failed hops; and conditional, contextual routing
of events. Consequently, events can be passed from agent to agent before
reaching their ultimate destination.
Mahout
Mahout is all about machine learning. The goal of the project is to build
scalable machine-learning libraries. The core of Apache Mahout is
implemented on top of Hadoop using MapReduce. However, the project
does not limit itself to that paradigm. At present, Mahout is focused on four
use cases:
•
Recommendation mining
: Recommendation mining is the driving
force behind several recommendation engines. How many of you have
seen something like this appear in your inbox: “Because you bought this
New England Patriots shirt, you might also like this NFL football.”
•
Clustering
: Clustering is the grouping of text documents to create
topically related groupings or categories.
•
Classification
: Classification algorithms sit on top of classified
documents and subsequently learn how to classify new documents. You
could imagine how recruitment agents would love clustering and
classification for their buzzword bingo analysis. If Apache Mahout is
able to reduce the number of calls received for the wrong job, that's a
win for everyone in my topic.
•
Frequent item set mining
: Frequent item set mining is a way to
understand which items are often bucketed together (for example, in
shopping basket analysis).
Ambari
Ambari is the system center of the Hadoop ecosystem. It provides all the
provisioning, operational insight, and management for Hadoop clusters.
Remember that Hadoop clusters can contain many hundreds or thousands
ofmachines.Keepingthemconfiguredcorrectlyisasignificantundertaking,
and so having some tooling in this space is absolutely essential.
Ambari provides a web interface for ease of management where you can
check on all the Hadoop services and core components. The same web
