WhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure service
IS OLAP DEAD IN THE AGE OF BIG DATA?
1. IS OLAP DEAD IN THE AGE OF BIG DATA?
AJAY ANAND, VP Products, Kyvos Insights Inc.
Yogesh Joshi, Head of Big Data and Analytics, AIG
2. AGENDA
• Why OLAP?
• Common implementation scenarios with Hadoop
• Issues with traditional tools connecting to Big Data
• OLAP on Hadoop: ROLAP, HOLAP, in-memory MOLAP,
distributed MOLAP on disk
• The Kyvos approach
• Use Cases
• Using OLAP for Risk Analysis
• Q&A
3. WHY OLAP?
• What OLAP can provide:
• Fast, interactive insights
• Ad hoc analysis
• Visual exploration, slice and dice, drill down
• Multi-dimensional view of data
• BUT, traditional OLAP solutions struggle with
• Massive increase in business data volume
• Explosion of cardinality (granularity) and dimensions
• Variety of data sources
4. WHAT DO USERS WANT?
• Self Service, Interactive Analytics
• On Big Data
• At any scale, with all kinds of datasets
• With instant response times, no waiting
5. BIG DATA SCENARIO
HADOOP
Cost effective
Scalable
Flexible
Hard to use
Not interactive
Transactional Data
Customer Interaction
Internet of Things
Web Interaction Logs
Product Usage Logs
Social Media Interactions
System Logs
Sales, Inventory, Revenue
Audit logs
Operational Data
Security Logs
Issues:
Accessibility, Ease of use,
Support for high
performance, interactive
analytics
Business Analyst
6. COMMON BIG DATA SCENARIO
Data Mart / EDW OLAP CubeHADOOP
Analytics /
Visualization Tool
Transform and Process on Hadoop,
Load into DBMS
• Transactional data
• Clickstream data
• Log files
• Other structured
and unstructured
data
7. ISSUES
Data Warehouse /
DBMS
OLAP Cube
HADOOP
Analytics /
Visualization Tool
Transform and Process on Hadoop,
Load into DBMS
• Transactional data
• Clickstream data
• Log files
• Other structured
and unstructured
data
SCALABILITY,
PERFORMANCE
LIMITATIONS
LATENCY
8. OLAP ON HADOOP
• Approaches
• ROLAP / HOLAP on Hive / Impala
• In-memory MOLAP
• MOLAP on Hadoop
9. KYVOS SOLUTION: MOLAP ON HADOOP
KYVOS
• Dashboards
• Interactive visualizations
• Explore, slice and dice, drill down
LOAD
HADOOP
CUBES ON HADOOP
TRANSFORM
• Transactional data
• Clickstream data
• Log files
• Other structured and
unstructured data
14. USE CASES
• Entravision:
• Consumer behavior analysis for Latino market
• Moving from sample and survey data (28K self reported diaries) to
empirical measurements on 15M+ adults
• Drill down to lowest levels of granularity
• Transactional data, purchase behavior, household characteristics,
consumer characteristics
• Precise measurability to prove efficiencies and ROI for media
planning
15. USE CASES
• Telecom subscriber profiling
• 20M subscribers
• 500B+ rows of data
• 60 days usage
• 96 node cluster
• Hourly incremental builds
16. USE CASES
• Risk analysis for financial services / insurance
• Operational analytics
• Web analytics for online shopping in the travel industry
• Set top data analytics