FusionInsight als Big-Data-Plattform - Eine Fallstudie aus der Telekombranche;
Dr. Walter Weigel, VP Huawei European Research Institute;
1st Smart Data Innovation Conference (SDIC'16)
SDIC'16 - FusionInsight als Big-Data-Plattform - Eine Fallstudie aus der Telekombranche
1. Huawei Technologies at Smart Data Innovation Lab, KIT
October 2016
Dr. Walter Weigel
VP of European Research Institute
2. Huawei is a leading ICT company
Carrier : 77% of Huawei's revenue
generated from the carrier network
business; from world's top 50 carriers
Enterprise: Serving more than 100 global
top 500 companies
Consumer : raising the brand awareness
to 65%
Leading global ICT solutions provider
A Fortune Global 500 company, ranking
285 in 2014
Interbrand Top 100 Best Global Brands
170,000+ employees worldwide
45% or 76,000+ employees engaged in
R&D
LinkedIn World's 100 Most InDemand
Employers
Who is Huawei
US$ 60B revenue in 2015
Serving 1/3 of the world's
population
Market Progress
Employees Business Areas
3. Telecom, a fast changing industry, what will be the future?
1980s
Voice
1990s
Voice/SMS
2000s
Data/HTML
2010s
Video
2020s
HD/VR/AR
2G
3G
4G 5G
1G
GSM
WCDMA
HSPA
LTE
LTE- Advanced
GPRS
2Mbps
10Mbps
100Mbps
1Gbps
10Gbps
100Kbps
10Kbps
??
4. IT, a fast changing industry, what will be the future?
1980s
Mainframes
1990s
Process
Computers
2000s
PCs, Servers
2010s
Cloud
platforms
2020s
HPCs
IBM
Evolution HPC &
Flash Memory
Siemens
Bull DEC
Apple
Nixdorf
HP
Compaq
GigaFLOP (1x109)
1984
TeraFLOP (1x1012 )
1999
PetaFLOP (1x1015 )
2008
ExaFLOP (1x1018 )
2019
ZetaFLOP? (1x1021 )
Beyond 2020
2.5 Tb/cm2
202050 Gb/cm2
20101 Gb/cm2
2000
5. 18 local R&D-sites in 8 countries with 1600 jobs
Ipswich
Dublin & Cork
Munich
Milan
Leuven
Paris
Nice
Gent
Nuremburg
Berlin
Gothenburg
Lund
Helsinki
Stockholm
Bristol
Cambridge
Huawei European Research Institute: Structure
Darmstadt
Establishment in Stockholm2000
Dec 2007
Apr 2009
Mar 2009
Dec 2008
Total People in Sweden 50+
Establishment of Bonn CentreMar 2008
Jun 2008 Establishment of Milan Branch
Major Movement from Bonn to Munich
Establishment of Gothenburg branch
Establishment of Belgium Branch
Jan 2012 Establishment of UK Branch
Sept 2011 Establishment of Nuremberg Branch
Dec 2012 Establishment of Finland and Ireland branches
Mar 2016
Acquisition of CaliopaAug.2013
Establishment of France BranchOct.,2013
Aestethics research FranceMarch 2015
Quantum Commun. Germany
6. 1600 researchers in Europe at 18 sites
Close to customers, close to industrial partners, close to academic partners
Goal: accelerate ICT innovations of future life and future work for a global market
Huawei European Research Institute: Overview
Telecom
IT
ITS / Mobility
E-/M-Health
Industries 4.0
Smart Grid
Smart City
Wireless Networks
Key Technical InnovationsIndustrial Solutions Fundamental R&D
Cloud and Big Data
Optical System
Terminals and IoT
+ +
New Theories
New Algorithms
New Materials
New Components
New Devices
7. Cooperation landscape in Europe (2015)
3+
5+
0.2+
31+
4+
10+
20+
5+
5+
2+
2+
0.8+
3+
0.7+
Unit: Million in Euros
Note: All projects are Huawei Funded.
8. Huawei joining SDIL
We are delighted to join the distinguished SDIL members
Intend to be a proactive member of the community, supporting research
Making hardware and the FusionInsight Big Data platform available for projects
Looking forward to working with the SDIL team, board members and with project
teams from corporations and universities, across all industries and areas of
research
Thank you for giving us this opportunity!
-------------------------------------------------------------------------------------------------------------------
A brief look at the FusionInsight Platform
An example of a recent Huawei project with Big Data
10. Big Data as a foundation for Innovation
HUAWEI Big Data PlatformDB / DW
Credit:
3~5 Weeks
Off Line
Historic Breakdown
Structured Data Structured, Semi-Structured, Unstructured Data
Credit
2~5 Seconds
On Line
Historic Breakdown
Possible
Assets
Financial
Networking
Targeted
Marketing
Service
Innovation
Traditional service …
11. FusionInsight Big Data Platform
BI Analysis
network signaling
analysis
CRM historical
data Inquiry
Data Collection
Real time data
BOSS
StructuredData
CRM
Other
System
……
Web
log
Unstructureddata
DPI
……
Internet
Data Backward
Web crawler
IDE & OM precision
marketing
Multi-tenant
Parallel Data Mining
(Customer Profile)
Streaming
Big Data Platform (FusionInsight)
Map Reduce
…… ……
X86 Server Cluster
detailed statement
inquiry
third party Data
Service
Data App
Short storage period, bad Customer satisfaction
High cost by RISC Server and SAN
…… ……
Difficulty with Scale-out
Performance bottleneck in inquiry
Challenges
Customer
Service
……
12. FusionInsight Big Data Platform
TelecomBank
FusionInsight
Enterprise
OceanStor
9000
FusionSphereFusionCubeX86 Server
Data insight: Parallel algorithms、Modeling、…
Off-line/Near-line
data
In-memory
iteration
Real-time
Processing
Manager
API API API APIPacking Services
Big data
infrastructure
Simple
• Life-cycle data management
• User-defined dashboard
• Redevelopment widgets
Reliable
• HA components & DR
• Distributed architecture
• N+M data protection
Real-time
• Real-time processing
• Industry-leading storage performance
Smart
• Deep insight based on full data
• Automated storage tiering
Supports 1000+km Disaster Recovery & Classified Protection of Financial Industry
14. Overview of Selected Telco Big Data Projects
Page 14
Business Services Objectives Huawei Key Offerings Clients
Churn management Reduce churn Predict churn accurately,
categorize churn reason
and propose solutions
CMCC, China Unicom,
Telkomsel,
Traffic management Increase traffic scale
and value
Drive traffic monetization
thru personalization and user
lifecycle value development
CMCC, China Unicom,
Mobily, Du, Vodafone,
PLDT, S.A. Telkom, …
Internet Operation Smart operation to
optimize, personalize
and monetize content,
services and user
experience
Big data enabled precise
marketing/sales solution to
optimize, personalize and
monetize data services
China Unicom
Intelligent location Monetize data assets B2B/B2B2C business
collaboration and trading
platform
CMCC, Facebook
16. Page 16HUAWEI TECHNOLOGIES CO., LTD.
Churn Management: Predictive Models for Churn Prevention
Can predict
likely churns
Know your customers.
Develop algorithms to
predict behavior.
Staying a step ahead
of the/each customer
is an important step in
preventing churn.
Potential
Churn
CUSTOMER VALUE MODEL
Input Data Algorithm Customer
Value (Grade)
CHURN PREDICTION MODEL
PRE-ALARM
INDEX
LIBRABY
Prediction
Algorithm
Pre-
alarm
Model
CHURN
MODEL
17. Promotion Sensitive
Tariff Sensitive
Product
Sensitive
Communication fee
promotion
On-net Age Promotion
Entertainment
News
Business
Life associated Service
Tools Service
Toll Discount Preference
Family and Friend
VPN Preference
Package Discount
1 VIP Service Club Score
Characters Segmentation Customer Needs
Amazing Service
Service
Sensitive
Service
Product
Price
Promotion
2
3
4
6
8
7
9
10 11
12 13
14
16
15 Terminal Promotion
Communication5
Insensitive Customer
Do not want to any promotion and service. They used to
consume regular tariff and service.
17Sloth
Churn: Customer Insight & Classification is essential
19. Churn reason Churn reason details
Category
Solution
Category Subcategory
Disappointment
churn
Operators seldom launch promotions. Promotion-sensitive Airtime promotion Top up promotion
Big promotions for new subscribers
rather than old subscribers
Promotion-sensitive
Promotion for staying with
operators
Promotion based on tenure
Cell phone losing and high fare for new
SIM application
Tariff-sensitive XXX Decrease fare for new SIM application
No personalized tariff plan or promotions
for individual users
Tariff-sensitive Low tariff preference Personalized tariff plans
Service-sensitive
Surprised service
preference
Service remaining and customer care
Competition churn Low tariff from competitors Tariff-sensitive Low tariff preference
Lease fare reduction
off-peak tariff reduction
Natural churn
Work location change Promotion-sensitive Airtime promotion Roaming promotions
Subscriber migrate to other place Promotion-sensitive
Promotion for staying with
operators
Keep phone number for a specific period.
Malice churn
Malice arrearage Tariff-sensitive
Tariff plan promotion
preference
High call fare remaining and arrearage alarm;
Promotion based on tenure
Agency channels encourage users to
churn and get commission
Promotion-sensitive Airtime promotion Delay commission payment to Agency channels.
Differentiating Reasons and identifying Solutions