As the big data market matures, Hadoop discussions are expanding from pure technology to how businesses can use it to drive innovation and leap-frog competition. In this session, Karmasphere will outline how to successfully deploy Hadoop projects by bringing together the right people, technology and use cases. We will discuss the optimal project team, the role of data scientists and analysts, the new big data analytics workflow and use cases for driving rapid ROI and success.
8. What Can We Make Happen?
“Data Platform of Intent”
Big Data
Full
Fidelity Analysis
Unstructured
Behavioral,
Open, Affordable
HADOOP
HQL
BI
Data Cube
Analysis
Structured, Transact
ional
Closed, Expensive
RDBS & EDW
SQL
Powered by the CommunityPowered by Vendors
Report On What Happened?
Database of Transactions
9. 1. What auxiliary products should we recommend?
1. What new features should our product have?
1. How can we eliminate support issues?
Hadoop Innovation Use Cases … some examples
10.
11. Another Idea
Score Your Predictive Models On Hadoop
Model Builder Model Description
Hive
UDFs
Standard
Hive
17. #1. Form a Partnership With LOB
Find a use case
Identify some budget
Form a project team
Be willing to educate others
Partner on a small POC, don’t boil the ocean
Hadoop Project Success
Best Practice #1
18. Teams are Highly Cross-Functional
Product Manager (LOB)
Power Analysts (IT or LOB)
Business Analyst (LOB)
Product Manager (LOB
IT Architect (IT)
Project Manager (IT/LOB)
19. “By 2016 the CMO will have
more budget than the CIO”
- Gartner Group
Marketing The “Budget Richest” LOB?
20.
21. #2. Use the Right Big Data Analytics tooling
• Supports the entire time
• Reuse and share for speed and efficiency
• Leverage pre-built analytics
Hadoop Project Success
Best Practice #2
23. #3. Embrace Full Fidelity Big Data Analytics
Not sampled, all the data – maintains richness
Don’t replicate or move the data
Keep complexity and TCO low
Hadoop Project Success
Best Practice #3
24. Data
Warehouse
OLTP to OLAP
Mapping
Analyst
In Summary
In BI, The Analyst Was at the End of the Process
Ordering App
Financial App
Master Data
Staging
OLAP
Reports
BI Using
Data Cube
Analysis
Structured, Sampled
Transitional, Closed
, Expensive
RDBS & EDW
SQL
Driven by Vendors
25. In Big Data Analytics on Hadoop
The Analyst is at the Center of the Process
Application
AnalyticsData
Unstructured
Behavioral,
Open, Affordable
HADOOP
HQL
Analyst
Full Fidelitiy
Analytics
Transform PMML models to standard Hadoop UDF’s. Leverage the power of Hadoop to score models across all data in Hadoop and increase their accuracy.
Today we are previewing Karmasphere 3.0 capabilities. One of the key new features we announced is the availability of a personalized Analytics dashboard. The dashboard helps users organize their big data analytics, presents the users with relevant analytics work products at their fingertips. It also provides activity status so you can complete projects on time and on budget