Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
Rethinking big dataHow big data and AI changing our lives
Pınar Uğurlu Kirazcı, Cloud Customer Engineer
Confidential & Proprietary
3
Tannat (BR, UY, AR) in cons truction
FASTER (US, J P, TW) 2016
Monet (US, BR) in cons truction for 2017
J unior (Rio, Santos ) in cons truction
PLCN Unity (HK, LA) in cons truction for 2018
FrankfurtBelgium
London
São Paulo
FinlandNetherlands
Hong Kong
3
Sydney3
Singapore
Sydney
Mumbai
Tokyo
TaiwanS CarolinaN Virginia
Oregon IowaMontreal
California
3
34
33
3
3
32
3 33
3
3
2
3
Edge points of pres ence (>100)
Future region and number of zones
Current region and number of zones
Google global cache edge nodes (>800)
Google leas ed/ owned fiber
GCP global infras tructure 15 Regions , 44 zones , over 100 PoPs , 100,000s miles of fiber optic cable
LANGUAGE API
VISIONAPI
APP ENGINE
COMPUTE ENGINE
KUBERNETES ENGINE
BIG QUERY
DATA FLOW
MACHINELEARNING
CLOUD STORAGE
NETWORKING COMPUTEENGINE
Teams, mobility, devices
Connected business platforms
App development & management
Data analytics & machine learning
Infrastructure, storage, network
Security / Scale / Control
Confidential & Proprietary
Competitive advantage ranked as top goal of machine-learning projects for 46% of IT leaders & 50% of adopters can quantify ROI
*Source: MIT Survey 2017; n=375Bain Consulting Study
2X more data-driven decisions
5X faster decisions
than others
3X faster execution
AI is the new ground for gaining competitive edge & creating bus ines s value
Confidential & Proprietary
If Your Company Isn’t Good at Analytics,
It’s Not Ready for AI“
”
“If your company is n’t good at analytics ,
it’s not ready for AI
– Harvard Business Review , 2017
Data challenges
Confidential & Proprietary* Harvard Bus ines s Review magazine; May-J une 2017
Less than 50% of structured data is used to make decisions*
Less than 1% of unstructured data is analyzed or used at all*
1< % < %50
Data analytics is s till too hard
Confidential & Proprietary
Focus on analytics not infrastructure
Develop comprehensive solutions
End-to-end ML lifecycle
Innovation and proven results
Our approach to data analytics
Leave scaling, performance, availability and security to Google Cloud’s serverless data analytics platform
Modern data warehouse, streaming data real-time analytics, advanced data visualization and AI
Operationalize predictive analytics as a logical next step in customers’ analytics journey
Proven track record of data analytics innovations. Leading enterprises rely on Google Cloud data analytics solutions
Confidential & Proprietary
Serverles s data analyticsFrom infras tructure to pla tform for ins ights
Performance tuning
Monitoring
ReliabilityDeployment & configuration
Utilization improvements
The traditional data analytics platform
Analysis and insights
Res ource provis ioning
Handling growing s cale
Analys is and ins ights
The s erverles s data analytics model
Confidential & Proprietary
Data ingestionat any scale
Reliable streaming data pipeline Advanced analyticsData warehousing
and data lake
Cloud Pub/Sub Dat a Transfer Service
Cloud Dat af low
Apache Beam
Cloud Dat aproc
BigQuery Cloud St orage Cloud MLEngine
Google Dat a St udio
Tensor f low
Complete foundation for data lifecycle
Sheet s
Confidential & Proprietary
Google papers
20082002 2004 2006 2010 2012 2014 2015
GFS MapReduce Flume Java
Opensource
2005
GoogleCloudproducts BigQuery Pub/ Sub Dataflow Bigtable
BigTable Dremel Spanner
ML
2016
Millwheel TensorflowDataflow
Fifteen years of tackling big data problems
Modernize your data warehous e foundation
Analyze s treaming data in real time
Be future-ready with AI
Get all your business data in one place for faster and comprehensive analysis
Gain real-time business insights and make your business more responsive
Simplify complex tasks with pre-learned machine learning engines
BigQuery: modernize your data warehous e Get a ll your bus ines s data in one place for fas ter and comprehens ive analys is
Confidential & Proprietary
Google BigQuery forms the AI foundation
Automate data delivery
Democratize data insights
Build the foundation for AI
Break data silos, power apps, add read-only data sets & make query results accessible to anyone
Automated data transfer to extract data from your systems & shared data with federated querying across any Google service
Enterprise Data Warehouse stores the most valuable data for your company & brings AI capabilities without replicating data into storage
Tee up real-time insights
Analyze real-time business events by automatic data ingestion, which is immediately available to query in your data warehouse
62 PetabytesLargest storage customer
BigQuery-scale performance
2.1 PetabytesLarges t query (data s ize)
4.5 Million rows/secPeak inges tion ra te
10.5 TrillionLarges t query (rows )
Confidential & Proprietary
Dataflow & Pub/sub:analyze s treamingdata in real timeGain real-time bus ines s ins ights andmake your bus ines s more respons ive
Confidential & Proprietary
Real time is real money
E-Commerce: Clicks tream analys is and dynamic user s egmentation
Retail: Proces s point-of-s a le transactions for real-time inventory pos itions
Mobile gaming: find the bes t Poké Ball collectors
Manufacturing: IoT data analys is for improving operational efficiency
Confidential & Proprietary
Ingest AnalyzeTransform
Cloud Dataflow
Machine learning & data warehouse
Ingest and distributedata reliably
Fast, correct computations quickly and simply
BigQueryCloud Machine
Learning
Cloud Natural Language API
Cloud Translation API
Cloud Vision API
Cloud Pub/Sub
Stream data analytics on Google Cloud Platform
20
Google Cloud : A Platform Ready for ML
Cloud AI Building Blocks
ML Professional Services
Language Conversat ionSight
Cloud Text- to-Speech
Cloud Speech-to-Text
Dialogf low Enterprise
Cloud Translation
Cloud Natural Language
AutoML Vision
Cloud Vision
Cloud Video Intelligence
Cloud Job Discovery
Contact center
Recommendation Engine
Cloud AI Solut ions
ASL Professional services organization
Cloud AI Plat f orm ML Librar ies
Tensorf lowCloud ML Engine
Cloud Dataproc
Cloud Dataf low
Datasets
Kaggle / Dat aset s
AutoML Natural Language
AutoML Translation
Kubef low
ML Hardware
Cloud TPUs Edge TPUs
Proprietary + Confidential
Sources: ComScore
Understanding s peech
cloud.google.com/ speech/cloud.google.com/ text-to-speech/
Proprietary + Confidential
Sources: ComScore
https:/ / research.googleblog.com/2016/09/a-neural-network-for-machine.html
Perfect translation
HumanNeural (GNMT)Phrase-based (PBMT)
English>
Spanish
English>
French
English>
Chinese
Spanish>
English
French>
Spanish
Chinese>
Spanish
Translation model
Tran
slat
ion
qual
ity
old: PBMT
new: GNMT
Understanding (other) languages
cloud.google.com/ trans la te/
Proprietary + Confidential
Machine Learning helped reduce error rates from 11% to 3% in the critica l proces s of correcting
s atellite image maps
Understanding images (cloud or s now?)
cloud.google.com/vision/
Proprietary + Confidential
Healthy Diseased
Hemorrhages
No DR Mild DR Moderate DR Severe DR Prolifera tive DR
Understanding dis eas es
Proprietary + Confidential
Understanding energy efficiency Google datacenters have half the
overhead (1.12 PUE) of typical indus try datacenters
Larges t private inves tor in renewables : $2 billion generating 3.2 GW
Applying Machine Learning produced 40% reduction in cooling energy
Proprietary + Confidential
Understanding defects
Proprietary + Confidential
Serverless, auto-s caling Data Pipeline for AI
Cloud Machine Learning
Proprietary + Confidential
Ingest data (s hock abs orber!)
Cloud Pub/ Sub
Images, meta data, gps, ... Cloud
Machine Learning
Proprietary + Confidential
Transform data (batch and s treaming)
Cloud Pub/ Sub
Images, meta data, gps, ...
Cloud Dataflow
Cloud Machine Learning
Proprietary + Confidential
Analyze data (Petabyte)
Cloud Pub/ Sub
Images, meta data, gps, ...
Cloud Dataflow
BigQueryCloud Machine Learning
Proprietary + Confidential
Train models (and us e them)
Cloud Pub/ Sub
Images, meta data, gps, ...
Cloud Dataflow
BigQueryCloud Machine Learning
Proprietary + Confidential
Train on your own images (AutoML Vis ion - no coding)
0.94 Defect
clean
defect
Confidential & Proprietary
DeveloperResponse within
4-8 bus ines s hours
The Forres ter Wave™: Ins ight Platforms -As -A-Service, Q3 2017.. The Forres ter Wave™ is copyrighted by Forres ter Res earch, Inc. Forres ter and Forres ter Wave™ are trademarks of Forres ter Res earch, Inc. The Forres ter Wave™ is a graphical repres entation of Forres ter's call on a market and is plotted us ing a detailed s preads heet with expos ed s cores , weightings , and comments . Forres ter does not endors e any vendor, product, or s ervice depicted in the Forres ter Wave. Information is bas ed on bes t available res ources . Opinions reflect judgment at the time and are s ubject to change.
“Our evaluation identified one vendor as a Leader based on the strength of its PaaS strategy, advancedtools for batch and real-time solutions, and machine learning and AI offerings.”
—The Forrester Report
● Google has the highes t s cores in the Current Offering and Strategy categories .
● Only vendor in the evaluation to offer ins ight execution features like full machine learning automation with hyperparameter tuning, container management, and API management.
● Receives recognition for advanced platform features like autos caling for mos t of its s ervices , efforts at integrating leading Hadoop cloud s ervices and its data flow s ervice works on both batch and s treaming data.
Google: a leader in ins ight platforms -as -a-s ervice
Rethink to Move your business forward on a s olid Google Cloud Big Data & ML foundationcloud.google.com/solutions/big-data
Teşekkürler