Transcript
Page 1: Cox Business Intelligence & Oracle BI/DW Stack

Cox Business Intelligence

& Oracle BI/DW Stack

John Landis, Sr. Director BI & Data Architecture

April, 2012

Page 2: Cox Business Intelligence & Oracle BI/DW Stack

2

Cox Communications Company Overview

• Is the third-largest cable

entertainment and broadband

services provider in the

country

• Has over 6 million customers

• Has over 22,000 employees

Page 3: Cox Business Intelligence & Oracle BI/DW Stack

3

Cox Communications Services

• Residential TV, Internet,

Phone, Tech Solutions, Home

Security

• Business TV, Internet, Phone,

Security, Backups, Industry

Services for Real Estate,

Residential Communities,

Education, Government,

Healthcare, and Hospitality

• Media On Air, Online and On-

the-Go

Page 4: Cox Business Intelligence & Oracle BI/DW Stack

4

CEBI 2008 Problem

• Business customers not

satisfied with multiple

platforms. Not sure where

to get data the right data.

• Business intelligence

platform has multiple

versions of the truth

• Data Integration is fractured

• Data Warehouse has not

had investment in 3 years

• Proliferation of tools has

become expensive and hard

to maintain

• Data needs of the company

are growing, offline

databases at all sites

• Development taking place in

multiple organizations

• No standards exist in the

enterprise

Page 5: Cox Business Intelligence & Oracle BI/DW Stack

5

Cox Enterprise Business Intelligence (CEBI) 2008

As well

as……

Page 6: Cox Business Intelligence & Oracle BI/DW Stack

6

CEBI 2008 Transition

Ad-Hoc Query

& Reporting

Standardized

Reporting

Advanced

Analytics

Modeling

Future Oriented

Operational Static

Strategic Dynamic

Past Oriented

Death by 1000 paper cuts

• Adopt a enterprise reporting

application to encourage

collaborative enterprise

development of reporting

across the organization and

lower the cost of reporting

throughout Cox.

• Reuse and optimization of

resources:

• People and Processes

• Application, Data, and

Services

• Time and Cost

Page 7: Cox Business Intelligence & Oracle BI/DW Stack

7

CEBI 2008 Solution

• Oracle Business Intelligence

Enterprise Edition 10G chosen

as the enterprise BI platform

• Oracle Database chosen as

the Enterprise data platform

and Infomatica chosen as

integration platform

• Business Intelligence

Competency Center Deployed

• Data Warehouse Clean-Up

Begins

Page 8: Cox Business Intelligence & Oracle BI/DW Stack

8

CEBI 2008 OBIEE 10G

• Total Cost of Ownership

• Common Semantic Layer

• Prebuilt Analytical Options

• Oracle’s Strategic BI Roadmap

• Single Sign On

• Embedded Metadata

• Self Service Reporting

Page 9: Cox Business Intelligence & Oracle BI/DW Stack

9

CEBI 2008 OBIEE 10G

• Interactive Charts and Graphs

• Personal Dashboards

• One Suite of Tools

• Open Source not ready for

enterprise deployment

• Hyperion Integration

• Personalization

Page 10: Cox Business Intelligence & Oracle BI/DW Stack

10

CEBI 2008 Solution Cont.

Cox lived happily ever after and I got to retire to my dream location…..

Not Exactly

Page 11: Cox Business Intelligence & Oracle BI/DW Stack

11

CEBI 2008 Lesson’s Learned

• Data governance is required

• IT can only facilitate data

governance, business needs

to lead

• Training is critical

• Self Service Reporting

requires supervision

• Start small

• Not everyone likes change

• Garbage in, garbage out

• Most people don’t understand

data, therefore carefully create

your RPD

• OBIEE resources are hard to

find

• The business wants you to

challenge them on requirements

• Reports are only as fast as the

database

Page 12: Cox Business Intelligence & Oracle BI/DW Stack

12

CEBI 2010 Exadata

• With the growing data and reporting needs within the

organization, the platform needed to expand to handle the

projected growth.

• Business data needs went from daily updates to near real

time updates.

• Existing hardware reached it’s capacity and new

technology was needed in order to meet the current and

upcoming demands.

• Without a platform and technology upgrade, data and

reporting would not be made available to the organization.

Page 13: Cox Business Intelligence & Oracle BI/DW Stack

13

BASELINEEXADATA AS

ISEXADATA NO

INDEXES

EXADATACOMPRESSI

ON/ AGGREMOVAL

EXADATAMIXED LOAD

TESTING

Total Time in Seconds 22,077 9,774 3,998 2,316 2,399

-

5,000

10,000

15,000

20,000

25,000

EXADATA POC -OBIEE Queries Total Run Time (Total of 130 queries executed) BASELINE

EXADATA ASIS

EXADATA NOINDEXES

EXADATACOMPRESSION/ AGGREMOVAL

EXADATAMIXED LOADTESTING

• In April 2010, the EBI team partnered with Oracle

to perform a Proof of Concept (POC).

• Based on the results of the POC, an executive

decision was made to implement the full solution.

• In July 2010 the EBI team began the planning and

rollout of Exadata.

• With the help of the Operations Support group, all

EBI databases were implemented on Exadata in

Production.

CEBI 2010 Exadata

Page 14: Cox Business Intelligence & Oracle BI/DW Stack

14

CEBI 2010 Exadata Pre-Launch Concerns

• People

- Support structure is different

- Adoption

- Learning curve for support and

development

• Process

- Compression can mask the lack of a

data lifecycle management

- It is not the way we have always

done it

- Performance increases can mask

architectural issues Note: Degradations in

performance caused by development code that should have

been avoided. Nearly 600K IOPS.

• Technology

- Oracle was new to the hardware market

- Technology had limited instances in

production.

- Switching from commodity based storage

to appliance; risk of stranding assets.

- Backup strategy changes and recovery

model changes

- Vendor “lock in” moving away from Oracle

becomes more expensive.

Page 15: Cox Business Intelligence & Oracle BI/DW Stack

15

CEBI 2010 Exadata

• Reporting

• Nearly 6X improvement out of the box

• Up to 200X query performance

improvement. 9X on average

• Nearly 6X performance increase on the

work orders load (non Exadata source).

2X on average for non Exadata sources

and 10X on average for Exadata to

Exadata loads.

• Some reports showed worse

performance

Page 16: Cox Business Intelligence & Oracle BI/DW Stack

16

CEBI 2010 Exadata Results

• 5-10x Compression saves Cox money over traditional storage

• Lowers backup time and tapes needed

• Estimated savings in space through 2012 range approx. $2.4M – $4.8M

Compression

• Less tuning reduces project timelines

• Enables near real time processing

• Able to process data previously not possible

• Estimated savings in time in 2012 approx. 5% or $400K

Performance

• Highly available, has uncovered issues in other Cox Oracle systems helping to improve reliability

• Reduces complexity of environment because Oracle has tested the integration points, all hardware is tested to work together unlike commodity solution

Enterprise Availability

• Oracle is our standard database today, no conversion costs were incurred.

• Cox employees already had a skill set in this technology

• Development best practices were enhanced

Leverage Existing Technologies

Page 17: Cox Business Intelligence & Oracle BI/DW Stack

17

CEBI 2012

Page 18: Cox Business Intelligence & Oracle BI/DW Stack

18

CEBI 2012 OBIEE 11G

• Score carding

• Mobility

• Improved Visualizations

• Spatial Intelligence via Map-

based Visualizations

• Business Process Invocation

• Packaged Apps

• Exalytics

Page 19: Cox Business Intelligence & Oracle BI/DW Stack

19

CEBI 2012 Architecture

Financial Data

Enterprise Metadata Layer

Financial

Consolidation

Planning, Budgeting

& ForecastingPROJECT

EMPLOYEE

CUSTOMER

ORGANIZATION

ERP -

FInancials

ScorecardsInteractive

Dashboards

Reporting and

Publishing

Adhoc

Analysis

Office

Integration

Mobile and

Embedded

Enterprise Business Intelligence Platform

Governance and Monitoring

Master Data

Detect and

Alerts

Collaborate &

Seach

OLTP Data

ERP - HR

CRM Logistics

Time &

AttendanceBilling

Data Warehouse

Human

ResourcesFinance

Sales &

MarketingField

Business

Operations

Reporting and Analytics

Data Sources

Pro

ce

ss

Sta

nd

ard

s

Page 20: Cox Business Intelligence & Oracle BI/DW Stack

20

CEBI 2012 Exadata

• Primary Database Areas:

- Reporting

- Applications

- Web Services

• Standby Database Areas:

- Analysis

- What If

- Predictions

- Data Mining

- Ad-Hoc

Page 21: Cox Business Intelligence & Oracle BI/DW Stack

21

Process Time 2008 – 10.5 Hours/Night 2009 – 12.5 Hours/Night 2010 – 4 Hours/Night 2011 – 3.5 Hours/Night

Data Volume 2008 – < 2 Billion/Night 2009 – 36+ Billion/Night 2010 – 43+ Billion/Night 2011 – 50+ Billion/Night

Users 2008 – < 1000 2009 – 2500+ 2010 – 5000+ 2011 – 9000+

BICC Migrations/Reviews 2008 – 100 2009 – 1300 2010 – 2218 2011 – 3000+

Errors Per 1M Sessions 2008 – 500 2009 – 150 2010– 125 2011– 100

User Generated Reports 2008 – <700 Usr Rpts 2009 – >6000 Usr Rpts 2010 – 10000+ Usr Rpts 2011 – 15000+ Usr Rpts

Complexity 2008 – Single Billing, Weekly/Nightly Numbers 2011 – Multi Billing, Near Real Time

CEBI 2012 History

Page 22: Cox Business Intelligence & Oracle BI/DW Stack

22

CEBI Customer Goals

Internal Cox Users want…

Product Planning and

Optimization Data Analysis and

Research

Real-time Operations

Monitoring

Data-driven Sales and

Marketing

Customers want…

Personal

Recommendations Product Personalization Customized Interfaces Personalized Services

Growing Analytical

Needs

Cross Product Usage Cross Team Efforts

Page 23: Cox Business Intelligence & Oracle BI/DW Stack

23

CEBI Future

Growth is taking place in areas not well served by traditional databases

According to Gartner, Enterprise Data will grow 650% by 2014. 80% of this data will

be Unstructured Data, with a CAGR of 62% per year, far larger than transactional

data.

The 2011 IDC Digital Universe Study Sponsored by EMC

This chart shows the growth over data over

the next couple of years. It is projected that

a large portion of this growth will be

unstructured data (web logs, emails, social

interactions, etc.).

Unstructured data does not work well with

traditional databases. To achieve the low

response times, traditional databases rely

on strict data structures. These strict data

structures work well for certain types of

data. However, the growth of unstructured

data in the enterprise and the proposed

uses of it create the need for a new type of

data processing to be introduced to the

technology stack.

Unstructured data is driving

an explosive growth in data

Structured

data

Unstructured Data in Web Pages

Page 24: Cox Business Intelligence & Oracle BI/DW Stack

24

CEBI Goals

Create additional value from customer data

Make Cox a more data-driven company

• Increase the perceived value of products by enabling a high degree of individual personalization.

• Give a highly tailored customer experience every time a customer interacts with Cox.

Democratize access to data

• Improve the efficiency and security of Cox operational processes.

• Allow the company to make decisions, spot trends, and react to competitive challenges more quickly.

• Allow the company can make quick, innovative use of the data that is already being generated every day.

• Improve cross-team and company wide insight into how customers are using Cox’s services.

Page 25: Cox Business Intelligence & Oracle BI/DW Stack

25

CEBI Goals

Cox Data

Structured Data

Characterized by well-known use cases, requiring only

repeatable, static data cubes and ETL’s. Highly

productized results.

Example Use Cases

• Ad-hoc Queries

• Financial and Operational Dashboards

• Ad Impression Analyzer

• Marketing Analyzer

Unstructured/Semi-structured Data

Characterized by lack of established use cases and on-

the-fly analysis in a “Sandbox” manner. Useful in

developing new insights, products.

Example Use Cases • Ad-hoc Queries

• Data Mining/Discovery

• Large Datasets, Fast Response Times

• Predictive Analysis

Traditional Data Architecture Big Data Architecture

Challenges

• Sizing to support new reporting dimensions is not

always economically feasible.

• Analysis against new datasets can slow Time to Market

for new products.

Challenges

•Latency is greater than with traditional databases.

• Large unstructured datasets will need to be monitored

and managed at scale.

Data Decision Framework

Page 26: Cox Business Intelligence & Oracle BI/DW Stack

26

CEBI Sample Decision Framework

• Used to evaluate analysis use

cases

• Can determine which system to

use:

• Traditional database

• Non-traditional data store

• Can standardize reporting and

analysis use cases across the

enterprise

Latency Complex Simple

High Exadata Any

Low Exadata Exadata

Latency Complex Simple

High Big Data Big Data

Low Big Data Exadata

High Data Volumes

Low Data Volumes

Page 27: Cox Business Intelligence & Oracle BI/DW Stack

27

Future Solution Design

Data Sources (Raw Data)

Stagie

Replication Virtualization Mediation Ingestion

Transfo

rm

Store

Presen

t Presentation Applications Analytics Services

Pro

cess

ODS Federated NOSQL Hadoop

MDM Virtualization ETL/ELT Map Reduce

Billing Mediation Anonymization Data Cleansing

EDW Virtual Master Store NOSQL

Page 28: Cox Business Intelligence & Oracle BI/DW Stack

28

POC in Progress

• Exalytics

• Oracle Big Data Appliance

• Endeca Information Discovery

• Packaged Apps


Recommended