Upload
others
View
5
Download
0
Embed Size (px)
Citation preview
AI
AI ⁻ Echo (hands-free
speaker) + Alexa (AI cloud )
⁻ iPhone + Siri
⁻ +Google Assistant
⁻ Microsoft Cortana
?:⁻
Echo
AI
ICT
,
: , ⁻ Tesla: 3 (2016 ).
. (Fremont )
⁻ Uber: CMU Robotics Institute
⁻ Google (220 ), Baidu:
TTesla: 3
Front TrunkApple iPad 17" Touch Panel
=> 70KW Supercharger 1 , 490km
( SW)(2016-6-30 )
BYD
Disruptive Innovation is verydifficult in big, market-leadingcompanies:
Strong boundary conditions oflegacy (product and service, people,partners,…).
Many University Startups atthis time of disruptiveinnovation
Industry 4.0, Platform 4.0Sense, Connect, Virtualize, & Servitize
Design Manufacture Assembly Test Delivery
Henning Kagermann (SAP CEO) .
Smart Factory E2E
Universities: Source of Innovation Startups
Quickstep Technologies(Aqcuired by Pivotal)
And SO many more….8
Apple: Alexa
Univ. of Washington(CMU) 2
Google DeepMind
Founded in September 2010,Acquired by Google in Jan 2014
vs in March2016
Google’s follow-up acquisitionsof Oxford university startups
Vision Factory (Computer Vision)Dark Blue Labs (Natural LanguageProcessing)
Cambridge Bill Gates ByuildingHall of Fame Awards
SAP HANA In-Memory Big Data PlatformRebellion was not easy!
SAP is the first (and old) software company founded in 1972Global market leader in enterprise software (ERP, CRM, PLM, Analytics, …)
SAP HANA Renewed SAP as Innovation Company >2X Market Value Growth (100B Euros) since SAP HANA announced in 2010SAP’s new platform for real-time digitalization of SAP’s enterprise customers and surrounding echo-system (e.g., collaboration with GE on Industrial IoT)Leading Industry Paradigm Shift
Hasso Plattner
A Crazy Professor’s Journey(2000 - 2014)
SNU Campus Startup in Seoul (2000)
EstablishTransact In Memory, Inc.
Silicon Valley (2002)
Interact with SAP Palo Alto Labs (since 2002)
Merger with SAP SE (Walldorf)
Establish SAP Labs Korea (2005) after SAP merger and Grow the team >10X
to Develop SAP HANAwith German Colleagues
Hyper (TU Munich Startup) acquired by Tableau (Nasdaq company from Stanford)
Thomas NeumannAlfons Kemper
In-memory database with data-centric code generation(LLVM)OLAP is serparted from OLTP using MVCCFully ACID-complient
Founded in 2013Fundings⁻ $13M from Andreessen
Horowitz⁻ $33M from New Enterprise
Associates, Andreessen Horowitz
Apache Spark
Ion StoicaFounding CEOCo-founder and CTO of ConvivaProf. of UC Berkeley
14
Matei ZahariaFounding CTOUC Berkeley Ph.D.Now Assist Prof at Stanford
: DJI
(HKUST) :
70%, 6
(2 )
2015
(~10% ) : US$ 8 Billion
2015 US$ 1 Billion
With camera, drones are intelligent data
acquisition and actuation devices of big data
systems as well as objects of autonomous
operation.
7 15 150 450 850
1,218 2,127
3,200
4,619
6,337
9,199
12,691
14,495
-
2,000
4,000
6,000
8,000
10,000
12,000
14,000
16,000
2005 2007 2009 2011 2013 2015 2017
Total Number of Employee (Facebook)
:10 100
2016.06.30
0
0
0
0
2016 : 1800 ( 2 )
=>
4 Big Data Governance Race: Time of Disruptive Innovation Everywhere⁻ M&A ⁻ Rule of “Minimum Five to Six” years from an idea to Potentially
Successful “Start Up”
R&D
“ ” R&D follower paradigm
4 :- . AI
- (Data Science and Innovation)
17
Suggestions for Digital-innovative Country
18
Political Independence of National R&D Governance:
Digital-innovative Human Resource Education
No more “Follower R&D”
Digital-innovative Transparent Government
Legal system for “Imagineering”
HANA HAUS in Palo Alto
19
In-Memory DB Technology Crossed the Chasm
In-Memory DBMS: Now in Main Stream
2010. 5.Announce HANA IM Platform
2011. 6.Release HANA IM Platform
2011. 9.Release BW on HANA
2013. 5.Release ERP on HANA
2012. 11.Announce IM OLTP HekatonRelease Plan
2013. 4.Announce DB2 IM BLU Release Plan
2013. 6.Release DB2 IM BLU
2013. 9.Announce IM 12c Release Plan
2014. 4.Release SQL Server IM OLTP Hekaton
2014. 6.Release Simple Finance on HANA
2014. 7.Release IM 12c
2012 2013 20142011
2013. 4.Release Cloudera Impala
Distributed IM
2013. 7.Release Apache Spark IM
2013. 9.Databricks Founded
2011. 6.HottonworksFounded
Open Source Hadoop Ecosystem
SAP HANA
Thomas Kuhn’s Paradigm Shift
(1922 – 1996)
“Science does not evolve gradually toward truth but undergoes periodic paradigm shifts”
Thomas Kuhnin the structure of scientific revolutions
Before the year 2010, no one believed that enterprise-scale in-memory database era is coming
soon (except a few in the world).
SAP HANA 1 - DBMS (1991-2000)2 : P*TIME (2000 – 2008): SAP (2005)SAP HANA (2009 - )
First Demo Proto with no name (2008.10) in SeoulAnnouncement (2010.5), First release (2011.6), BW porting (2012), CRM & ERP porting (2013)
Node 1
Node NNode 2
Data+log
Data+log
Data+log
Company[CHAR50]
Region[CHAR30]
Group[CHAR5]
< Row Table > < Columnar Table >
The world-first 1TB-DRAM blade ready for SAP HANA
nodes (2009.4Q)
(16GBX8=128GB)
6TB Memory on 60-Core Box (2014)
4 Socket, 60-core, Intel Ivy Bridge
6TB DRAM (96 X 64GB DIMM)
SAP HANA ERP Paradigm Shiftfrom 3-tier to 2-tier with computation down where data resides
S/4 HANA: SAP is rewriting ERP Suite to fully leverage the real-time analytics power of SAP HANA’s in-memory columnar database
26
SAP HANA ERP Paradigm Shift
S/4 HANA DB : - DBRedundant
-(SAP HANA , )
( : ): ( )
( )
EMR
2KHzSensor
200HzSensor
IoTHow can we extend this experiment to smart factory?
KPI ()
Universities: Source of Innovation Startups
Quickstep Technologies(Aqcuired by Pivotal)
And SO many more….31
Don’t forget: Google was a University StartupStanford Professor Gio Wiederhold at DARPA created US Digital Library Project together with others in NSF and NASA in 1993At Stanford, Prof. Hector Garcia-Molina led the project.Page Ranking Paper (1996), Google Founded (1998)
Gio Wiederhold
Hector Garcia-Molina
Michael Stonebraker: Turing Award Winner (2014)
Professor at UC Berkeley (1971-2000), MIT (2001-present)Numerous Commercialization of University Research Prototypes: Ingres, Postgres (Illustra), C-Store (Vertica), StreamBase, H-Store, VoltDB, SciDB and many other projects
Test-of-Time Talk in IEEE ICDE 2015 Seoul
“Big 4” bound by “Innovator’s Dilemma”
:
DARPA ( ), NSF( ) R&D ,
Google : IPO 11⁻ 1993 DARPA, NSF, NASA Stanford Digital Library
, 1998 , , 2004 IPO Search Engine Cloud, Big
Data, AI,
: , Google, Apple M&A
: Biomedical Data Science
:
Meds & LabsClinicRecord
GenomicPatterns
Complex Analysis Process
Medical Records
PersonalGenomic
Characteristics
: 1
⁻ : , , , ,
⁻ : 2025
⁻ (2020 60% )
⁻ : , , R&D
⁻ track record R&D 3 100
41
67
132
150
214
264
1.1
7.1 7.0
4.9
0
5
10
15
20
25
30
-
50
100
150
200
250
2011 2012 2013 2014 2015 2016.11.04
()
)
Tencent
Tencent (WeChat )
2014-10-01,
(6 )
2012-04 ,720
(Bill
ion
US D
olla
rs)
4.8 ( 2400 )
: ( )
Andrew Ng (AI Stanford
)
R&D
Google AI project
Coursera (MOOC 1 )
Deep Learning , /
Merck: 348
Exec Board Member
" 4.0 ”
" , "
:?
5 ,
19 R&D⁻
?
⁻ ?
?
, ?
:?
Fast Follower , ?
Governance : ⁻ 1
?
?
? ⁻
SSNU Big Data Institute:
Established in April 2014 with de facto national status 10% of SNU faculty (200 professors) in all disciplines including social and natural sciences, medicine, and engineering.Global Network with Japan (Tokyo, Kyoto), China (Tsinghua, Shenzhen CAS), Germany (TU Berlin, Dresden), Stanford, etc.Ongoing Discussion on Establishing a University-wide Graduate School of Data Science Innovation
Trans-disciplinary Problem SolvingTelco CDR and IPTV Log Data (with Korea Telecom)Precision Medicine with SNU Hospitals with the world-first In-Memory Clinical Data Warehouse (based on SAP HANA)Urban Data Science topics: Transportation, Environment, Healthcare of Aging people (on UDS Innovation Campus provided by Seoul City Government)Smart Cars and Self-Driving CarsSmart Energy Grid Big Data ...
Market Value Changes Bloomberg’s Ten Largest Tech Companies
$662b
$521b
$433b
$168b $74b
$138b
$162b
$299b$190b
$134b
Yellow: positive gainsRed : negative
Jan 16 => Nov 25
What commonalities can welearn from blue and red ones?
Industry in Big Data Governance Race
Ex: Google, Facebook, Microsoft, Apple, GE, …: Those who govern big data will govern industryOn-going structural change is so broad and fast: called “4th
Industrial Revolution”Companies which cannot transform themselves (overcome the innovator’s dilemma) will lose market power (i.e., market value)Conventional corporate R&D is slow to change, and also is limited in attracting new talents: Acquire startups to survive.Start-ups prosper, especially from university research – New Paradigm of University Research Transfer to Real World!
44
An Outlier Challenging Self-Driving Car
George Hotz, 26-year-old hacker built a self-driving car
with deep learningCapability in One Month
Shenzhen
46
Conclusion
Exciting era of big data researchers⁻ Industry in Big Data Governance Race⁻ Big and Fast Structural Changes in Almost All Industrial Sectors
“Build, Measure, and Analyze” (= BMW from Mike Carey)⁻ Understand Technology (Limitations) & Market (Problems)Avoid Herd Mentality⁻ Historically not all hyped research created impact
Look for Dimensions of Innovations⁻ New domains, or existing domains with 10X changes (Andy Grove’s criteria) in hardware
and software along levels of computing abstraction⁻ Essentially involves thinking about which boundary conditions can be
removed/relaxed how far“Don’t worry about failures”: Passion will bring successMoney should not be the primary goal: It will follow
Global Big 4
High Competition in “Advanced” In-Memory DBMS
Total market share > 88%Only 2.4% sales growth
(11-14 Oracle CAGR)No chance to new faces
Startups
Tech-centric venturesOften using open source
Hadoop ecosystemsOver 80% sales growth (11-14 Cloudera CAGR)
Variety of chance
Global Big Data Market Trend
48
49
AI AI AI AI Service Big Datavirtuous circle
AI Service and Big Data
Servitization: GE case
50
GE CEO Jeff Immelt:
“We'd like to be at $10 billion by 2020”“On track to be a top 10 software company”
virtuous circle
IoT Service Big Data AnalyticsImprovements
Sensor Data