Weitere ähnliche Inhalte Ähnlich wie Real-Time Data Integration for Modern BI (20) Kürzlich hochgeladen (20) Real-Time Data Integration for Modern BI1. Real-time Data Integration for Modern BI
Claudia Imhoff
Intelligent Solutions, Inc.
Boulder BI Brain Trust
Jake Freivald
Information Builders
2. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Claudia Imhoff
2
President
Intelligent Solutions, Inc.
Founder
Boulder BI Brain Trust
A thought leader, visionary, and practitioner, Claudia
Imhoff, Ph.D., is an internationally recognized expert
on analytics, business intelligence, and the
architectures to support these initiatives. Dr. Imhoff
has co-authored five books on these subjects and
writes articles (totaling more than 150) for technical
and business magazines.
She is also the Founder of the Boulder BI Brain Trust,
a consortium of internationally recognized
independent analysts and experts. You can follow
them on Twitter at #BBBT or become a subscriber at
www.bbbt.us.
Email: claudia@bbbt.us
Phone: 303-444-6650
Twitter: Claudia_Imhoff
3. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Agenda
Introduction to Operational BI
The Extended Data Warehouse Architecture
Putting It All Together – Examples
Getting Started with Operational BI
3
4. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
What is Operational BI?
A set of applications, services, and technologies
for monitoring, reporting on, analyzing, and
managing the business performance of an
organization’s daily business operations
4
5. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
The Business Case
Enables more informed business decisions by directly
supporting specific business processes and activities
Supports faster business decisions by seamlessly
integrating BI with business processes to create a
closed-loop decision-making environment
Provides a more dynamic business environment where
the business can learn, adapt, and evolve based on
analysis of its operational business performance
5
6. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Agenda
Introduction to Operational BI
The Extended Data Warehouse Architecture
Putting It All Together – Examples
Getting Started with Operational BI
6
7. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Next Generation – Extended Data
Warehouse Architecture (XDW)
Traditional EDW
environment
Investigative computing
platform
Data
refinery
Data integration
platform
Analytic tools & applications
Operational real-time environment
RT analysis
platform
Other internal & external
structured & multi-structured
data
Real-time streaming data
Operational systems
BI services
7
8. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Data Provisioning Use Case:
Data Integration
8
Heavy lifting process of extracting,
transforming to standard format
and loading structured data –
mostly batch
Physically consolidates data into
“trusted” EDW sets for analysis
Invokes data quality processing
where needed
Employs low-cost hardware and
software to enable large data
volumes to be combined and stored
Requires more formal governance
policies to manage data security,
privacy, quality, archiving and
destruction
Traditional EDW
environment
Investigative computing
platform
Data
refinery
Data integration
platform
9. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Data Provisioning Use Case:
Data Refinery
9
Ingests raw detailed structured and
unstructured data in batch and/or
real-time into a managed data store
Distills data into useful business
information and distributes the
results to downstream systems
May also directly analyze certain
types of data
Also employs low-cost hardware
and software to enable large
amounts of detailed data to be
managed cost effectively
Requires (flexible) governance
policies to manage data security,
privacy, quality, archiving and
destruction
Traditional EDW
environment
Investigative computing
platform
Data
refinery
Data integration
platform
10. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Traditional EDW Use Cases
10
Most BI environments today
New technologies can be
incorporated into EDW
environment to improve
performance, efficiency & reduce
costs
Use cases
Production reporting
Historical comparisons
Customer analysis (next best
offer, segmentation,
life-time value scores,
churn analysis, etc.)
KPI calculations
Profitability analysis
Forecasting
Traditional EDW
environment
Data
refinery
Data integration
platform
Analytic tools & applications
Operational real-time environment
RT analysis
engine
Operational systems
BI services
11. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Investigative Computing Use
Cases
New technologies used here
include:
Hadoop, in-memory
computing, columnar storage,
data compression,
appliances, etc.
Use cases
Data mining and predictive
modeling for EDW and real-
time environments
Cause and effect analysis
Data exploration (“Did this
ever happen?” “How often?”)
Pattern analysis
General, unplanned
investigations of data
11
Data
refinery
Data integration
platform
Analytic tools & applications
Operational real-time environment
Investigative computing
platform
RT analysis
engine
Operational systems
BI services
12. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Operational BI Use Cases
Embedded or callable BI
services:
Real-time fraud detection
Real-time loan risk
assessment
Optimizing online promotions
Location-based offers
Contact center optimization
Supply chain optimization
Real-time analysis engine (Event analytics):
Traffic flow optimization
Web event analysis
Natural resource exploration analysis
Stock trading analysis
Risk analysis
Correlation of unrelated data streams
(e.g., weather effects on product sales)
12
Operational real-time environment
RT analysis
platform
Other internal & external
structured & multi-structured
data
Real-time streaming data
Operational systems
BI services
13. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Operational
BI
Operational BI Environment
Consists of:
RT Event
Analytics
Investigative
Analytics
13
Embedded BI
Services
14. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Agenda
Introduction to Operational BI
The Extended Data Warehouse Architecture
Putting It All Together – Examples
Getting Started with Operational BI
14
15. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
All Components Must Work
Together
15
analytic models
other analytics
New sources of data
Enterprise DW
Analytic tools
Investigative
computing
platform
Data refinery
Operational
systems
production
analyses
next best
customer offer
location-based offer
fraud detection
3rd party data
location data
social data
feedback
RT analysis engine
call center dashboard
or web event stream
NOTE: data virtualization has a big
role in the combining of analytics
16. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Example – Online Promotions
Real-time campaigns
Adjust pricing and offers every
few minutes based on
inventory levels and customer
responses
Faster reaction to supply-
demand imbalances
Immediate customer metrics
New customer, retention,
repeat buys
Average revenue, profit, etc.
Faster operational
(merchandizing) reports
16
17. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Example – Location-based
Offers
Empower local stores to do own special events and
promotions
Non-technical, in-store personnel must be able to select
customer groups based on local criteria
Appropriate reminder and incentive offers
Exclusive invitations, coupons, etc., pushed to mobile
devices based on consumer history, proximity to store,
external events
17
18. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Example – Algorithmic Trading
Used in capital markets
to indicate when to buy or
sell stocks
Applies event processing
by calculating complex
algorithms on the fly
Example: When GM’s
price is .5% higher than
average price in last 30
seconds, buy 10,000 share
of Ford every 3 seconds
until average price drops
18
19. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Example – Retail Supply Chain
Automation
Use RFID to automate
supply chain
Can track, trace items
wherever they are, at any
time
RFID-sourced events must
be quickly collected,
organized, managed by
data management
infrastructure
Event data volumes may
not be large but event
rates are fast
19
20. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Example – Fraud Detection
Idea is to detect fraudulent
transactions faster
Credit card transactions
Telecom transactions
Claims
Models are developed from
known fraudulent events in
data warehouse and fed to
streaming analytics engine
Human can intercede,
eliminate proven fraudulent
transactions quickly
20
21. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Other Operational BI Examples
Web-site analytics
Loan and credit card application processing
Insurance claim processing
Improving airline customer satisfaction (flight
delays, baggage handling, call-center handling)
Monitoring and optimizing equipment servicing
and quality management (IoT)
Package shipment, routing, and delivery
optimization
Network outage tracking, prediction & correction
21
22. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Agenda
Introduction to Operational BI
The Extended Data Warehouse Architecture
Putting It All Together – Examples
Getting Started with Operational BI
22
23. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Barriers and Challenges*
23
* From: tdwi.org/research/2009/09/ten-mistakes-to-avoid-when-designing-and-developing-operational-bi-applications.aspx?tc=page0
24. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Does Your Business Need
Real-Time BI?
Understand what the requirements for real time delivery of
data means to the business community
What MUST be real time?
What CAN be minutes to hours old?
What CAN be days, weeks old?
How important is real-time access and analytics to your
company – what is the ROI for real-time processing?
Remember costs and complexity go up as need for real
time data and analysis increases
24
25. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Cost Considerations: Operational
BI is Mission Critical
Sound infrastructure
A robust data integration
architecture (data analytics)
An integrated BI and business
process environment
(streaming analytics)
A services-based operational
BI infrastructure
Support for mixed workloads
Data analytics from DW and
experimental areas
Streaming analytics from RT
analytics engine
Reliable service levels – High
availability/fault tolerance
25
26. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Assess Reality – 1
ID time continuum for data in operational BI processes
Assess existing data delivery capabilities
Understand business operational BI requirements
What is right time for accessing data?
Use a time continuum (from real-time to high latency) to map
business needs to type and cost of data delivery approach used
26
Traditional query
Time
At-rest data
(past events)
Start
time End
time
processes existing data
processes real-time event streams
Stream query
In-motion data
(continuous events)
27. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Assess Reality – 2
Understand which weaknesses will be
exaggerated as you speed up analytical and
operational processes
Weak data quality processes
Weak ETL processes
Weak workflow integration
Weak adherence to procedures by employees
27
28. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Pick a Project - 1
Two options:
Improve function and/or speed of existing BI solution
Build a new operational BI solution that solves a new
or important business need
1. Improved BI solution
More sophisticated analytics,
e.g., predictive analytics fed to operational personnel
Reduce time – may be evolutionary process based
on experience, e.g., daily to intra-day to near real-
time data and analyses
28
29. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Pick a Project - 2
2. New BI solution
Recognized as having significant impact on
operations but achievable in reasonable time frame,
e.g., 3 to 6 months
May fill a gap in current operations, e.g., fraud
detection, location-based offers
May be made possible by advances in BI technology,
e.g., event analytics, extreme analytic DBMS
platform
29
30. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Pick a Project - 3
Look for operational workflow activities that
have a major impact on costs or revenues
Bottlenecks or issues that can be made more
efficient through the use of operational BI
Understanding business processes & costs of
activities within business processes is a critical
success factor
30
31. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Pick a Project - 4
Avoid making big changes
to operational processes
Operational BI is an
evolutionary process
Focus on improving the
speed and efficiency of
existing processes
Realize you may have to
modify existing standard
operational policies and
procedures
May need alerts if procedures
are not followed
31
32. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Operational BI: Users are
Different
Traditional BI supports a few
100 (or maybe a 1,000) users
Opening BI up to operational
personnel means ramping up
to many more users
May increase both software
and hardware costs
Means tighter, faster
connectivity of enterprise
decision support environment
to rest of company
32
33. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Operational BI: Dashboard
Example
Mechanism to present analytics, alerts,
recommendations, decision workflows, operational KPIs
Used to help make proactive decisions
Send alerts and recommendations to other personnel
Ability to drill down to further understand what’s happening
Must be self-contained and easy to use – personalized
to each manager’s needs
Self-service – Business person can change metrics, thresholds
being tracked; presentation and layout of results
Easy to modify as business scenarios and needs evolve
Level of changes allowed will depend on role of user
33
34. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Operational Impact
Workforce may need to be revamped – will
likely require retraining of operational personnel
How they make decisions
How they access and use BI information
How they monitor impact of their decisions
Training needs to be ongoing and flexible
Some may not make the leap
34
35. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Presentation Requirements – 1
Continuous availability of
operational data and BI results
Current information from
operational systems
Integrated with BI analytics on
demand
Minimal impact on operational
systems performance
Presented in proactive manner
May need dynamic modeling
Business person may change
business rules on the fly
Show different set of metrics
depending on situation
35
36. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Presentation Requirements – 2
Transparent underlying infrastructure
Study their access methods and needs
Develop appropriate dashboards, portals, or other interfaces
according to these needs
Monitor community’s usage patterns and revamp, revise the
interface as needed
Workbench based on workflow
Bring together appropriate historical BI results, BI services,
streaming events, operational capabilities to support workflows
Consider data virtualization to minimize data movement, simplify
changes
Remain flexible – things will change!
36
37. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Infrastructure Requirements - 1
Operational BI is mission critical
Both operational and BI
infrastructure may need
upgrading to meet service-level
agreements
BI performance (# of users,
response times, volumes of data,
types of analytics performed)
Availability, reliability, security
Impact on operational processing
Need accurate predictions of
data and user growth
37
38. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Infrastructure Requirements - 2
Solution should have flexible
(elastic) scalability, query support
and capacity
Support for workload
management & mixed workflows
essential
POC required to ensure all parts
work
Data integration and data quality
will be different
Sometimes “good enough” data
is good enough …
Other times, data must be
perfect!
38
39. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Data Integration Main
Considerations
Consider:
Source data type, organization, and scale
Data quality requirements and techniques
Source data transformation & validation
rules
Target data type, organization, scale,
latency, granularity
Data integration techniques & technologies
Data integration approaches for data
analytics
Data virtualization – connect BI insights
to operational data
Change data capture (CDC)
Data staging: ELT, streaming data &
Hadoop filtering
Data integration infrastructure is THE
cornerstone of operational BI
39
40. Copyright © 2016, Intelligent Solutions, Inc. All Rights Reserved
Operational BI: 10 Mistakes to
Avoid
1. Assuming All Analytics Must Come from the Data Warehouse Alone
2. Failing to Match BI Agility to Business Needs
3. Failing to Determine if the Infrastructure Can Support Operational BI
4. Assuming that Operational BI is Just a Technology Solution
5. Assuming that Operational BI Simply Involves Capturing More Timely Data
6. Assuming Existing Data Quality Procedures Will Work for Operational BI
7. Failing to Realize Operational BI is Process-Centric Rather than Data-Centric
8. Assuming Operational BI IT Skills are the Same as Those for Other BI Types
9. Assuming Users of Operational BI are Same as Those for Other Types of BI
10.Failing to Monitor and Audit Automated Decision Making
From TDWI’s 10 Mistakes to Avoid When Designing and Developing Operational BI
Applications: tdwi.org/research/2009/09/ten-mistakes-to-avoid-when-designing-and-
developing-operational-bi-applications.aspx?tc=page0
40
42. The Information Builders 3i Capabilities
Portal Embedded InfoApps™
ApplicationsLegacy Systems Relational/Cubes Big Data Columnar/In Memory Unstructured Social Media Web Services Trading Partners
Integration
Mobile Write-Back
Data
Discovery
Reporting Dashboards
High-Performance
Data Store
Data
Quality
Data
Governanc
e
Master Data
Management
Batch
ETL
Real-Time
ESB
Integrity
Intelligence
Location
Analytics
Casting
and Archiving
In-Document
Analytics
SearchPredictive
Analytics
Sentiment and
Word Analytics
Performance
Management
Social
Hot
Bad
Feedback
45. The Power of Information Builders
Create operational insights, analytical content
…and drive them immediately to 7x-20x more users
45
Interoperable integration, intelligence, and integrity
46. The Power of Information Builders
Create operational insights, analytical content
…and drive them immediately to 7x-20x more users
Create predictive models
…and use them to score data in motion (e.g., into Hadoop)
46
Interoperable integration, intelligence, and integrity
47. The Power of Information Builders
Create operational insights, analytical content
…and drive them immediately to 7x-20x more users
Create predictive models
…and use them to score data in motion (e.g., into Hadoop)
Leverage a Real-time Data Quality Firewall
47
Interoperable integration, intelligence, and integrity
48. The Power of Information Builders
Create operational insights, analytical content
…and drive them immediately to 7x-20x more users
Create predictive models
…and use them to score data in motion (e.g., into Hadoop)
Leverage a Real-time Data Quality Firewall
Virtualize data across operational data stores
48
Interoperable integration, intelligence, and integrity
49. The Power of Information Builders
Create operational insights, analytical content
…and drive them immediately to 7x-20x more users
Create predictive models
…and use them to score data in motion (e.g., into Hadoop)
Leverage a Real-time Data Quality Firewall
Virtualize data across operational data stores
Drive information directly to users in real-time
Example…
49
Interoperable integration, intelligence, and integrity
51. Sample Architecture Benefits
Event driven by input stream, real real-time
Application content centrally managed by iSM
Information is pushed to the consumer, not pulled
Application logic not necessary in browser
Scalable: a single object can be pushed to all users, no
handling of AJAX requests per browser session
Reusability: output content stream available for any
application, not just browsers
Based on accepted open standard protocols
51
52. Sample Architecture Benefits
Event driven by input stream, real real-time
Application content centrally managed by iSM
Information is pushed to the consumer, not pulled
Application logic not necessary in browser
Scalable: a single object can be pushed to all users, no
handling of AJAX requests per browser session
Reusability: output content stream available for any
application, not just browsers
Based on accepted open standard protocols
52
53. Hadoop in Operational BI
Generate data sources for business users / exploration
Not everything needs to be modeled
E.g., 1/2 to 1/3 of customer data lives in a data
warehouse
How do you manage analytics that involves other data?
Traditional: DW
DWApp
New
Traditional: Op’l
DWApp
New
Modern
DWApp
New
Traditional: Rogue
DWApp
New
54. Integrating Big Data
54
Data at rest and data in motion
Sqoop,Flume…
Avro,JSON…
Traditional
applications
and data stores
iWay Hadoop Data Manager
Simplified, modern,
native Hadoop integration
Big Data Hadoop
Any distribution, Any data
BI &
Analytics WebFOCUS BI and analytics platform
Self-service for Everyone
WebFOCUS access,
ETL, metadata
WebFOCUS access,
ETL, metadata
55. Simplified
interface
Native Hadoop
script generation
Process mgmt
& governance
Simplified, easy-to-use interface
to integrate in Hadoop
Marshals Hadoop resources
and standards
Takes advantage of performance
and resource negotiation
Includes sophisticated process
management & governance
Sqoop,Flume…
Avro,JSON…
Data Sources
Big Data Native: Runs in Hadoop cluster
Purpose-built: Exploits all Hadoop services
Simple: Replaces coding with mapping
Integrating Big Data
55
Data at rest and data in motion