Contexti / Oracle - Big Data : From Pilot to Production

Copyright © 2014 Oracle and/or its affiliates. All rights reserved. |
Big Data: From Pilot to Production
Vicky Falconer -Oracle
Grant Priestley -Contexti

my.oracle.com/go/bigdata
Program Agenda
1
2
3
4
Big Data Project Challenges
Typical Big Data Journey
Common Operating Models
Technology Considerations

Big Data Project Challenges

Experience with a regional Telco
•You don’t know what you don’t know…
•Build it and they will come
•Technology versus capability
•Clear definition of skill requirements
•Moving from Pilot to Production

Additional challenges
•Who owns data?
•What to do in house and what to externalise
–Analytics
–Admin
–Development
–Engineering
•Operationalisinginsight

Key Lessons
•Where to start?
•Culture
•Scope
•Building capabilities
•Technology
Right Questions
Right Use Cases
Right Business Case

Typical Big Data Journey

Big Data Developments
Increasing use of infrastructure as a service
Better understanding of the possibilities offered by unstructured data
Moving from historical batch computing to real-time analytics
Wider awareness and a more defined understanding of Big Data
Wider variety of vendors offering Big Data solutions
Less hype, more real use cases of companies exploiting Big Data
Maturity of Big Data tools bringing them into the mainstream
What changes we have noticed over the past 12 months with respects to Big Data that are most likely to impact on your organisationor on the market in general:
Increase in requests for platform as a service
Advanced
Less Advanced
2015?

Where Value Lies for Most Organisations
The proliferation of Big Data Analytics applications and solutions has given rise to the need for a Big Data Platform that enable these initiatives to occur and support all use cases including Advanced Analytics, Internet of Things (IoT) and the Digital Enterprise. The Big Data PaaSaccelerates organisation'sprojects by provisioning the initial platform and development environment, eliminating the need for hard-to-find Big Data skills and ultimately allows the enterprise to focus on strategic initiatives and IP creation rather than platform operations.

Big Data Use Cases
1
2
3
Customer Insights / Behavior
Data Warehouse Augmentation
Risk Analysis & Fraud Detection

Customer Insight / Behavior
Challenges
Understand customer behavior through predictiveanalytics
Solution
Leverage Hadoop and ML techniques to build “population-based behavioral” clusters enabling personalisedcontent to be served up in certain real-time sequences
BusinessOutcomes
•Increase in sales conversion
•Online engagementis personalised, as it is in store

my.oracle.com/go/bigdata Copyright © 2014 Oracle and/or its affiliates. All rights reserved. |
Data Source(s) Contexti Big Data Platform Data Consumer(s)
Customer Insight / Behavior
Semi-Structured
Data
Structured
Data
Pre-computed
Web Content &
Deals
Raw / Enriched
Data Sets
(HDFS / MFS)
Streaming
Data
Acquisition
File-Based
Data
Acquisition
RDBMS-Based
Data
Acquisition
Data
Ingestion
Deep Analytics
&
Machine
Learning
Streaming Capability Serving Capability
Batch Capability
Online Store
Customers

Challenges
Reducelatency between data generation and availability
Solution
OffloadETL processing to Hadoop platform and support the ingestion of multi-structured data sets
BusinessOutcomes
•Access ofdata reduced from T+1,T+2 to real-time / intra-day
•Reduce cost of ETL processing
•More time now spent on analysingdata than data wrangling

my.oracle.com/go/bigdata Copyright © 2014 Oracle and/or its affiliates. All rights reserved. |
Data Source(s) Contexti Big Data Platform Data Consumer(s)
Semi-Structured
Data
Unstructured
Data
Raw / Enriched
Data Sets
(HDFS / MFS)
Streaming
Data
Acquisition
File-Based
Data
Acquisition
RDBMS-Based
Data
Acquisition
Data
Ingestion Extract Load
Transform (ELT)
and Data
Preparation
Processes
Streaming Capability Serving Capability
Batch Capability
Users
Structured
Data
Reporting,
Search &
Query
RDBMS &
MPP
Platforms
Pre-computed
Views

Challenges
Reduce incidentsof fraud through more sophisticated detection and monitoring
Solution
Ingestedstructured and raw law data from multiple applications and combined data filtering from Pig/Hive with statistical modeling by R,whileexecuting CEP on streams of data
BusinessOutcomes
•Implementionof real-time trigger based analytics that provides early detection of fraud
•“Schema on read” provided greater flexibility for analysis

Contexti Big Data Platform
Data Consumer(s)
Data Source(s)
Semi-Structured
Data
Unstructured
Data
Raw / Enriched
Data Sets
(HDFS / MFS)
Streaming
Data Acquisition
File-Based
Data Acquisition
RDBMS-Based Data Acquisition
Data Ingestion
Streaming Capability
Serving Capability
Batch Capability
Online Store
Structured
Data
Data Access Provisioning API
RDBMS & Analytics
Platforms
Raw Data
(In-Memory)
CEP / Stream Analytics
Pre-computed Views
Real-Time Incremental Views
Fraud Systems
Deep Analytics &
Machine Learning

Decentralised
Centralised
Federated
In a decentralized services model, each business or function has its own analytics group, which enables and encourages rapid decision-making and execution.
Pros:
•Analytics needs aligned tobusiness functions
•Close to business and customers needs
Cons:
•Limited strategic view
•Duplication, redundancies and inability to standardise or leverage scale
The centralised shared-services model exists outside organizational divisions or functions, in some cases externalto the organisation itself.
Pros:
•Standardised processes and methods
•Independentviewpoints shared across the organisation
Cons:
•Perception that group lacks functional expertise
•Ownership of IP when outsourced
The federated shared-services model is a centralized model that rolls under an existing function or business unit and serves the entire organization.
Pros:
•Speedin execution & decision making
•Pre-existing shared service processes and structure
Cons:
•Less transparent resource allocation
•Focus on business function priorities

Building a Best Practice Analytics Capability
Analytical Capability
Techniques
Questions
Basic
Provide static, historical view of business performance drawn on basic scorecard andstatic reports
Queryand drill down
Where is the problem?
Adhoc reporting
How many?How often? Where?
Standardreporting
What happened?
Anticipatory
Creates transparency into pastand future drivers, using systems and processes to perform a range of descriptive analytics
Segmentation Analysis
Whatare the unique drivers?
Statistical Analysis
Why is this happening?
Sensitivity Analysis
What if conditions change?
Predictive
Requires high-qualityintegrated data and complex mathematical capabilities and offers dynamic forward looking insights
Optimisation
What is the best that can happen?
Simulation
What would happen if …?
Predictive Modelling
Whatcould happen next?

Challenges to Face
Hurdles between Pilot and Production

Pilot vsProduction Characteristics
Pilot Characteristics
Production Characteristics
Funding
Project-Based
Project and BAU Funding
Numberof Use Cases
2–3 use cases
> 5 use cases
Insights
Demonstrated
Actionable/ Operational
Service Level
No / Loose SLA (project-based)
EnforcedSLAs, OLA
BigData Capability
•Batch
•Serve
•Batch
•Serve
•Stream (advanced)
Resiliency/ DR
No
Mandatory
Security Enabled
Optional
Mandatory
Scale
1 Rack, <5 data sources
Multiple Racks, >10 data sources
Timing
3-6 months
6-9 months*

Is Your Big Data Pilot Ready for Production?
Culture
•Allows for trial and error, ability to fail
•Understanding that data is an enterprise asset, benefit of being “data informed”
Structure & Skills
•Governance of data and its use
•Decision on what skills to acquire/develop/buy(analyst, dev, data scientist, ops, engineering)
•Funding Model (How will users/customers be charged?)
Integration
•Technologies in place to connect internal/external data, including unstructured data
•Integration of “actionable insights” into operational processes

•Business Strategy drives IT Strategy
–Information Architecture
•Future State Infrastructure
–Scale out and up
–Adding Big Data to existing infrastructure can be complex
•Analytics
–Embed in operational systems
•Integration insight into existing systems and processes

ActionableEvents
Streaming Engine
Data Reservoir
Enterprise Data & Reporting
Discovery Lab
Actionable
Information
ActionableData Sets
Input
Events
Execution
Innovation
Discovery Output
Data
Conceptual View
StructuredEnterprise
Data

Oracle Big Data Strategy
BY INDUSTRY & LINE OF BUSINESS
BIG DATA APPLICATIONS
DISCOVERY
BUSINESSANALYTICS
BUSINESS ANALYTICS
DATA RESERVOIR
BIG DATAMANAGEMENT
DATA WAREHOUSE
SOURCES

Oracle Big Data Management System
SOURCES
DATA RESERVOIR
DATA WAREHOUSE
Oracle Database
Oracle IndustryModels
Oracle Advanced Analytics
Oracle Spatial & Graph
Big Data Appliance
Apache Flume
Oracle
GoldenGate
Oracle Event Processing
Cloudera Hadoop
Oracle NoSQL
Oracle R Advanced Analytics for Hadoop
Oracle R Distribution
Oracle Database
In-Memory, Multi-tenant
Oracle Industry Models
Oracle Advanced Analytics
Oracle Spatial & Graph
Exadata
OracleGoldenGate
Oracle EventProcessing
Oracle DataIntegrator
Oracle Big DataConnectors
Oracle DataIntegrator
ORACLE BIG DATA SQL

Oracle R Enterprise Approach
Data and statistical analysis are stored and runin- database
Same R user experience & same R clients
Embed in operational systems
Complements Oracle Data MiningROpen Source

Closing Remarks

Summary
•Where to start?
•Culture
•Scope
–Data Ownership
–Data Governance
–IM Strategy
•Building capabilities
•Technology
Right Questions
Right Use Cases
Right Business Case

Contexti / Oracle - Big Data : From Pilot to Production

Recommended

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to Contexti / Oracle - Big Data : From Pilot to Production (20)

Recently uploaded (20)

Contexti / Oracle - Big Data : From Pilot to Production