DRIVING INTELLIGENCE TO THE EDGE – FEATURING AN END-TO …
Transcript of DRIVING INTELLIGENCE TO THE EDGE – FEATURING AN END-TO …
1
DRIVING INTELLIGENCE TO THE EDGE – FEATURING AN END-TO-END OPEN ARCHITECTURE FOR IoT
Strata Data Conference, September 12, 2018 Dave Shuman, Cloudera, Industry Lead for Manufacturing and IoT Bryan Dean, Red Hat, Director Business Development for IoT Solutions
2
Securely connect, authenticate and manage disparate connected devices that speak different protocols
Apply analytics at the edge with machine learning and business rules to enable local, low-latency decision making
Centralize IoT data processing, analytics and machine learning to enable deep business insights and actionable intelligence
Create and deliver cloud-native applications. Integrate business applications and processes.
Tools to enable end-to-end data security, compliance, authorization and authentication
KEY FUNCTIONALITY FOR AN END-TO-END IoT ARCHITECTURE
Device Management & Connectivity
Intelligent Edge Processing & Analytics
Advanced Analytics & Machine Learning
Business & Application Integration
End-to-End Security & Compliance
3
WHY OPEN SOURCE FOR IoT
“We believe the best way to support this complex environment is to base our commercial IoT platform, the Bosch IoT Suite, on open source components and open standards. These projects establish a horizontal open technology for IoT and provide the technical breeding grounds for successful business ecosystems.” - Dr. Stefan Ferber, VP of Engineering, Bosch Software Innovations
2.4 30* 250+ 130K
million lines of code
projects developers monthly visitors
4
ADDRESSING ENTERPRISE IoT NEEDS
Data Management & Analytics
● Enterprise Data Mgmt.
● Persistent Data Storage
● Big Data Processing & Analytics
● Real-Time Analytics
● Machine Learning
● Data Security & Compliance
Operational Technology (OT)
● Device Management
● Industrial protocols
● OT Middleware
● Intelligent gateways
● MQTT co-inventors
● OT security
Information Technology (IT)
● Messaging & Integration
● Business Rules & CEP
● Open Hybrid Platform-as-a-Service
● Enterprise Linux Platform
● IT security
Enterprise IoT open source community
Operational Technology (OT) Information Technology (IT)
5
public, private, hybrid cloud
DATAMANAGEMENT&ANALYTICSPLATFORM
IoTEDGE
CONNECTED“THINGS”
IoTINTEGRATIONHUB
OPEN END-TO-END IoT ARCHITECTURE
APPLICATIONDEVELOPMENT,DELIVERY,&INTEGRATION
Integrating IoT operating technology, data management, analytics, and applications
Cloud-native apps
Traditional apps
• Modular, secure, end-to-end architecture • Streaming analytics and machine learning • Open, interoperable on hybrid cloud • Modern application development and agile integration
6
CONNECTED“THINGS”
Sensors, Actuators,
Data Sources
Edge Processing & Analytics
DATAMANAGEMENT&ANALYTICSPLATFORM
IoTINTEGRATIONHUB
APPLICATIONDEVELOPMENT,DELIVERY,&INTEGRATION
Data Integration, Routing, Device
Command/Control Application Development, Deployment, Integration
Advanced Analytics & Machine Learning
EDGE PROCESSING AND ANALYTICS
IoTEDGE
• Device connectivity • Data transformation • Intelligent routing • Edge applications • Edge analytics & real-time
decisions
7
Edge Processing & Analytics
Data Integration, Routing, Device
Command/Control
TelemetryData
Application Development, Deployment, Integration
Advanced Analytics & Machine Learning
DATA INTEGRATION, ROUTING, DEVICE COMMAND/CONTROL
IoTINTEGRATIONHUB
CONNECTED“THINGS”
Sensors, Actuators,
Data Sources
DATAMANAGEMENT&ANALYTICSPLATFORM
• Device management, security, access control
• Data aggregation • Event processing • Agile integration • Container-based
application platform
IoTEDGE
APPLICATIONDEVELOPMENT,DELIVERY,&INTEGRATION
8
Edge Processing & Analytics
Data Integration, Routing, Device
Command/Control Advanced Analytics & Machine Learning
Application Development, Deployment, Integration
• Data ingest • Stream / batch processing • Secure data storage • Machine learning and
real-time analytics
IoTEDGE
ADVANCED ANALYTICS AND MACHINE LEARNING
Machine Learning Model
CONNECTED“THINGS”
Sensors, Actuators,
Data Sources
DATAMANAGEMENT&ANALYTICSPLATFORM
IoTINTEGRATIONHUB
APPLICATIONDEVELOPMENT,DELIVERY,&INTEGRATION
Telemetry Data
9
Application Data
Advanced Analytics & Machine Learning
APPLICATION DEVELOPMENT, DEPLOYMENT, INTEGRATION
Application Development, Deployment, Integration
Cloud-native apps
Traditional apps
APPLICATIONDEVELOPMENT,DELIVERY,&INTEGRATION
Edge Processing & Analytics
IoTEDGE
CONNECTED“THINGS”
Sensors, Actuators,
Data Sources
IoTINTEGRATIONHUB
Data Integration, Routing, Device
Command/Control
DATAMANAGEMENT&ANALYTICSPLATFORM
• Container-based application platform with Kubernetes orchestration
• DevOps: automated build-test-deploy • Agile integration • Multi-cloud portability
10
IoTEDGE
CONNECTED“THINGS”
IoTINTEGRATIONHUB
DATAMANAGEMENT&ANALYTICSPLATFORM
Cloud-native apps
Traditional apps
OPEN END-TO-END IoT ARCHITECTURE
• Device connectivity • Data transformation • Intelligent routing • Edge applications • Edge analytics & real-time
decisions
• Data ingest • Stream / batch processing • Secure data storage • Machine learning and
real-time analytics
• Container-based application platform with Kubernetes orchestration
• DevOps: automated build-test-deploy • Agile integration • Multi-cloud portability
APPLICATIONDEVELOPMENT,DELIVERY,&INTEGRATION
• Device management, security, access control
• Data aggregation • Event processing • Agile integration • Container-based platform
11
IoTEDGE
CONNECTED“THINGS”
IoTINTEGRATIONHUB
DATAMANAGEMENT&ANALYTICSPLATFORM
Cloud-native apps
Traditional apps
OPEN END-TO-END IoT ARCHITECTURE: PRODUCTS
APPLICATIONDEVELOPMENT,DELIVERY,&INTEGRATION
ENTERPRISEDATAHUB
12
IoTEDGE
IoTINTEGRATIONHUB
CONNECTED“THINGS”
Protocol Translation
IntelligentFiltering Aggregation Routing
DATAMANAGEMENT&ANALYTICSPLATFORM
Real-Time Analytics
Data Ingest Real-Time Processing
Data Storage
Machine Learning
Data Security
TelemetryData
Application Integration
Management
Data flow to derive deep business insights and actionable intelligence END-TO-END ANALYTICS
Telemetry Data
Deepdataanalysis&insights
Application Data
Cloud-native apps
Traditional apps
APPLICATIONDEVELOPMENT,DELIVERY,&INTEGRATION
13
Protocol Translation
IntelligentFiltering Aggregation Routing
Deepdataanalysis&insights
Real-Time Analytics
Data Ingest Real-Time Processing
Data Storage
Machine Learning
Data Security
Telemetry Data
Application Integration
Application Data
Data flow to derive deep business insights and actionable intelligence END-TO-END ANALYTICS
Machine Learning
MLModel
Actions
Prediction/Alert
MLModel
IoTINTEGRATIONHUB
DATAMANAGEMENT&ANALYTICSPLATFORM
Cloud-native apps
Traditional apps
CONNECTED“THINGS”
IoTEDGE
APPLICATIONDEVELOPMENT,DELIVERY,&INTEGRATION
14 Red Hat Partner Confidential
PUTTING IT INTO PRODUCTION
Data Mgt. and Analytics
/home/cdsw/trained-model.h5
IoT Edge
IoT Hub
Connected “Things”
Predict Locally
Capture Images
Download Model
Report Result
A trained model becomes a “downloadable brain”
16
END-TO-END IOT ARCHITECTURE - PoC
Application Data
Telemetry Data
MQTT over WiFi
Telemetry Data
Command & Control
Telemetry Data
Application Integration
Client XDK Data Mgt. and Analytics
IoT Edge
Network: Red Zone Network: Green Zone
Telemetry Data
IoT Integration Hub
Applications
Digital Twin
17
IoT Integration Hub
Application Data
Telemetry Data
MQTT over WiFi
OT Middleware OT Middleware
Smart Services
Machine Learning
Business Logic
Device Management
Device Connectivity
Administration
Platform-as-a-Service
Ingest
Machine Learning
Store
Analyze
Process
Telemetry Data
Command & Control
Telemetry Data
Application Integration
Client XDK
Digital Twin
Data Mgt. and Analytics
BI IoT Edge
Network: Red Zone Network: Green Zone
Apps CDSW PPM
Telemetry Data
END-TO-END IOT ARCHITECTURE - PoC
18
OPC-UA
Kafka Spark Kudu
Integrate Store
Spark
Impala
Analyze
Unified Services Layer
YARN Sentry
Data Science Workbench
TCP
/IP
Mqt
t / O
PC
-U
A
IoT Calibrators Edge IoT Site Hub
TCP/IP mqtt
• Device connectivity • Data transformation • Intelligent routing • Business logic • Edge analytics & real-time
decisions
• Device management, security, and access control
• Data aggregation • Event processing • Integration services (API’s)
TCP/IP Kafka
Model Deployment
DATA PIPELINE & ARCHITECTURE
• Data ingest • Stream / batch processing • Persistent data storage • Machine learning and
real-time analytics
20
Stacks for Machine Learning at the Core and Edge
CORE EDGE
Machine Learning Library
Machine Learning Platform
Operating System
Cloud / On-Prem Infrastructure
Model Training Application
Machine Learning Library
Edge Software Library / Framework
Operating System
Physical Device
Model Serving Application
21
So what about a Classifier?
CORE EDGE
DL4J
Cloudera Data Science Workbench
RHEL
AWS EC2 & S3
TrainClassifier.java
DL4J
Everyware Software Framework (Java OSGi)
Yocto Linux
Eurotech ReliaGATE 20-25
ServeClassifier.java
Machine Learning Library
Machine Learning Platform
Operating System
Infrastructure
Model Training Application
22
Or…
CORE EDGE
Keras (w/ TensorFlow backend)
Cloudera Data Science Workbench
RHEL
Azure & ADLS
train_classifier.py
Tensorflow Lite (Java API)
Everyware Software Framework (Java OSGi)
Yocto Linux
HPE Moonshot
ServeClassifier.java
Machine Learning Library
Machine Learning Platform
Operating System
Infrastructure
Model Training Application
23
TECHNICAL OVERVIEW: FUNCTIONALITY
IoT EDGE
CONNECTED “THINGS”
Telemetry
Management
IoT INTEGRATION HUB
Machine learning model
APPLICATION DEVELOPMENT, DELIVERY, & INTEGRATION
CONTAINER-BASED APPLICATION PLATFORM
Traditional applications Cloud-native applications
ERP
CRM
DBs
DATA MANAGEMENT & ANALYTICS
Data engineering
Data science
Data warehouse
Operational database
Security & governance
Machine learning
Storage Services
ENTERPRISE DATA HUB
Telemetry
Management
App integration
CONTAINER-BASED APPLICATION PLATFORM
IOT EDGE FRAMEWORK
DevOps
Integration &
API mgmt Orchestration
Developer services
Edge analytics
Machine learning
Field connectivity
Cloud connectivity
Edge applications
Telemetry
IOT INTEGRATION FRAMEWORK
Device management
Device connectivity
Data integration
25
TO LEARN MORE…
Visit the Cloudera Booth
Industry 4.0 Demo
Featuring the End-to-end architecture and solution
Join us for a Webinar on Oct 3rd
to hear about the end-to-end Solution
28 Red Hat Partner Confidential
PREDOMINANT END-TO-END IoT ALTERNATIVES
Alternative Strength Weakness
Full-Stack Public Cloud Providers
• Quick to get started • Low initial investment • Offload in-house expertise • Single vendor source/support
• Loss of data control • Proprietary lock-in, limited portability • Rigid architecture • Limitations at edge for many use cases • Cost model at scale
Proprietary IoT Platforms
• Single vendor packaging • Medium-quick to get started • Optional vertical use-case packages
• Proprietary lock-in, poor portability • Black-box architecture • Rigid architecture • Requires vendor-specific in-house knowledge • Limited interoperability
Do It Yourself • Perceived low initial software cost • Customized to use case • No vendor middleman
• Vast in-house expertise required • Slow time to value; high development and ongoing
costs • Support and indemnification • Complexity, maintainability
29
TECHNICAL OVERVIEW: FUNCTIONALITY
IoT EDGE
CONNECTED “THINGS”
Telemetry
Management
IoT INTEGRATION HUB
Machine learning model
APPLICATION DEVELOPMENT, DELIVERY, & INTEGRATION
CONTAINER-BASED APPLICATION PLATFORM
Traditional applications Cloud-native applications
ERP
CRM
DBs
DATA MANAGEMENT & ANALYTICS
Data engineering
Data science
Data warehouse
Operational database
Security & governance
Machine learning
Storage Services
ENTERPRISE DATA HUB
Telemetry
Management
App integration
CONTAINER-BASED APPLICATION PLATFORM
IOT EDGE FRAMEWORK
DevOps
Integration &
API mgmt Orchestration
Developer services
Edge analytics
Machine learning
Field connectivity
Cloud connectivity
Edge applications
Telemetry
IOT INTEGRATION FRAMEWORK
Device management
Device connectivity
Data integration
30
TECHNICAL OVERVIEW: PRODUCTS
IoT EDGE
CONNECTED “THINGS”
Telemetry
Management
IoT INTEGRATION HUB
Machine learning model
APPLICATION DEVELOPMENT, DELIVERY, & INTEGRATION
Telemetry
Management
App integration
Telemetry
Edge analytics
Machine learning
DATA MANAGEMENT & ANALYTICS
ENTERPRISEDATAHUB
31
TECHNICAL OVERVIEW: PROJECTS
IoT EDGE
CONNECTED “THINGS”
Telemetry
Management
IoT INTEGRATION HUB
Machine learning model
APPLICATION DEVELOPMENT, DELIVERY, & INTEGRATION
DATA MANAGEMENT & ANALYTICS Telemetry
Management
App integration
Edge analytics
Machine learning
Telemetry
CLOUDERA’SDISTRIBUTIONINCLUDINGHADOOP(CDH)
32
IoT Gateways
IoT Integration Hub
REMOTE MAINTENANCE & SUPPORT
ON-SITE DATA ANALYTICS
MACHINE LEARNING & ADVANCED ANALYTICS
Data Mgmt, Analytics & ML
Manufacturing Equipment
● Data ingest ● Stream / batch
processing ● Secure data storage ● Machine learning and
real-time analytics
● Device connectivity ● Data transformation ● Intelligent routing ● Business logic ● Edge analytics &
real-time decisions
● Device management, security, and access control
● Data aggregation ● Event processing ● Integration services
INDUSTRY 4.0 DEMO ARCHITECTURE
33
VALUE PROPOSITION
Open and interoperable Future-proof open source architecture | open standards | deployment flexibility
Modular Avoid lock-in | capitalize on existing investments
End-to-end analytics Analytics at the edge | advanced analytics and machine learning | ML model execution at the edge
Reduce risk and complexity
Simplify development, deployment, and integration tasks | save costs
End-to-end security Security across devices, access, authentication, and applications as well as data in motion and at rest
Control your data Privacy | security | regulatory
34
Device connectivity Open standards – MQTT, AMQP, OPC-UA, CoAP, HTTP(s)
Flexible deployment
Data management & analytics Based on Apache open source ecosystem libraries for machine learning and advanced analytics
Open application interfaces
Enterprise visibility | real-time anomaly detection | future-proof
Community innovation Collaboration driven by some of the leading enterprises in the IoT space
Any of the leading cloud providers or your data center or hybrid cloud
No vendor lock-in No rigid architectures or proprietary formats and components
OPEN SOURCE, OPEN STANDARDS, FLEXIBLE DEPLOYMENT