EDGE PATTERNS IN THE IIOT
BRAD NICHOLAS
CHICAGO IOT MEETUP
MARCH 2017
2Copyright © 2017 Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx
AGENDA
01 3 minutes about Uptake
02 Some key considerations
03 The 3 patterns
04 Manufacturing discussion / Q&A
3Copyright © 2017 Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx
Uptake at a glance
AEROSPACE AGRICULTURE CONSTRUCTION ENERGY
104M
predictions
generated to date
2014
founded in Chicago
82%
across Data Science
& Engineering
700 Employees
Uptake has developed partnerships in:
HEALTHCARE MINING RAIL RETAIL
Uptake selected as the hottest
startup of 2015 – beating out
Uber and Slack. – Dec 2015
Uptake’s Industry Thought Leaders featured in:
4Copyright © 2017 Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx
Our platform is purpose-built to deliver actionable insights and
recommendations into workflow, empowering people to create value
Raw Data Data Ingestion Platform Apps
Data Science Engines
Data Integrity
Software Development Kit
Failure Prediction
Anomaly Detection
Recommendations
Event / Alert Filtering
Data Operations Center
Normalization &
Cleansing
End to end visibility
Encryption in transit
and at rest
API Portal
Developer
Content
Mgmt.
App Store
Tools
Assets
Customers
ERP
Contextual
• Weather
• Social Media
• 3rd party
Sample Apps:
• Condition-
Based
Monitoring
• Supply Chain
Optimization
• Fuel and
Energy
Management
• Performance
Optimization
Workflow Integration
Examples:
• Automated
locomotive
re-routing
• Automated
parts
ordering
• Automated
maintenance
scheduling
END-TO-END CYBER, INFORMATION, AND OPERATIONAL SECURITY
5Copyright © 2017 Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx
About Me
I run the IoT team at Uptake
bradn
www.linkedin.com/in/bradn
Automotive, Manufacturing,
Consulting, Telecom, Startups
EE MBA
Fun fact: I “OEM+” hack & restore
German cars
6Copyright © 2017 Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx
We’re hiring.
https://boards.greenhouse.io/uptake
Come see me if you’re interested in IoT, device management, embedded programming, crypto
Key Considerations
02
8Copyright © 2017 Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx
Digitization is lagging in many industry sectors that need IIoT
MGI Industry Digitization Index
http://www.mckinsey.com/industries/high-tech/our-insights/digital-america-a-tale-of-the-haves-and-have-mores
• Quasi-public and/or highly localized
sectors are lagging in digitization
• Labor-intensive sectors need digital
tools for the workforce
• Knowledge-intensive sectors are
already highly digitized
• Capital-intensive sectors have high
IoT potential
• Service sectors can digitize customer
transactions
• B2B sectors can benefit from
expanded digital engagement
6
5
1
2
3
4
9Copyright © 2017 Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx
3 essential elements to IIoT value creation
Data Ingestion
“Sense”
Analytics
“Infer”
Workflow
“Act”
10Copyright © 2017 Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx
IIC’s reference model for industrial analytics covers most of the bases
Multi-tiered approach
Sensing vs Actuating
Different time horizons
Open vs Closed loop
Source: Industrial Internet Consortium IIRA http://www.iiconsortium.org/IIRA.htm
11Copyright © 2017 Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx
Where you compute affects many things
There is no one architecture that will address everything.
But there are certainly some common questions to answer
Proximity
Response
Time
Node
Computing
Capacity
Bandwidth
Consumed
Focal
Points
Exceptions
Sense Act
12Copyright © 2017 Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx
How you are able to connect also affects what you can do
Latency, bandwidth, cost and complexity are usually not as optimal as you want them to be
MobileLocal IndividualSite
13Copyright © 2017 Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx
Other key IIoT needs, beyond strong security & viable economics
Separation of Concerns is essential
Key to managing complexity, achieving
maintainability and resilience
https://effectivesoftwaredesign.com/2012/02/05/separation-of-concerns/
IP protection is crucial
Data rights management for both original
and derived data, at rest and in flight, all
nodes, including authorized use
https://motherboard.vice.com/en_us/article/why-american-farmers-are-
hacking-their-tractors-with-ukrainian-firmware
Heterogeneity is unavoidable
Computing environments
Node state
Mobility vs fixed location
Networking options and node availability
Domain responsibility
IT/OT barrier is literally a real thing
Operational control comes first
Skills/expertise is very different
Most capital equipment is decades old and
relies on physical security
http://blog.iiconsortium.org/2016/08/it-vs-ot-for-the-industrial-internet-two-
sides-of-the-same-coin.html
The 3 patterns
03
15Copyright © 2017 Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx
3 patterns seem to address most IIoT deployment scenarios
Physical Edge
On-device IoT node
Platform &
Applications
Cloud Edge
reverse CDN for
the physical web
Edge Gateway
“On location”
connectivity node
16Copyright © 2017 Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx
The Cloud Edge is effectively the ‘virtual physical web’
A hybrid node that serves as a “concentrator” or “reverse CDN” for the physical web.
It can isolate IoT traffic and service cloud-based applications with anything they need from the
physical web
Concentrates physical web data streams
Interacts with Edge Gateways and higher end Physical Edge nodes
Serves web APIs to cloud applications
You can train ML using the data on this node.
You could continually train ML given sufficient compute capacity and data.
You can distribute its contents via CDN, subject to data rights management
17Copyright © 2017 Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx
The Physical Edge interacts directly with IIoT data sources
Protects the OT layer and hosts specialized, “high interaction” IoT processes
Serves as a direct data extraction point for physical web data generated by a machine or process
Protects machine / process operation at all costs, even if data extraction compromised
Runs on-machine / on-process analytics functions
Protects OEM and machine owner IP by enforcing data rights management at the source, under terms
suitable to the IP owners
Must be designed and deployed in collaboration with machine / process OEMs and operators
Provides much richer data access capabilities
18Copyright © 2017 Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx
The Edge Gateway
Resides in proximity to physical web nodes and handles connectivity gaps
Manages “inter physical web” IoT interactions that aren’t needed to control things
Primary function is to monitor physical web machines / processes
Eliminates the need for physical web devices to interact with the Cloud Edge directly
Queues on premise when backhaul connectivity is unavailable, restricted due to cost or otherwise unusable
Speaks local machine dialects
19Copyright © 2017 Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx
The 3 edge patterns can be implemented flexibly
Physical Edge nodes deployed on advanced machines with excellent connectivity can connect
directly to Cloud Edge nodes – without an Edge Gateway
Cloud Edge nodes could be deployed anywhere connectivity to other edge nodes and “data
center quality” bandwidth is available
• A very high end physical web machine or process
• At a fixed location like an airport terminal
Edge Gateway nodes could be co-deployed with Physical Web nodes as long as suitable
backhaul connectivity to a Cloud Edge node is available
Discussion
Applying the edge patterns in manufacturing
04
21Copyright © 2017 Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx
OEE - Overall Equipment Effectiveness
Total Productive Maintenance
Seiichi Nakajima 1982-1984
www.AMTonline.org
http://capstonemetrics.com/files/whitepaper-oeeoverview.pdf
OEE = Availability x Performance x Quality
TEEP = Loading x OEE
22Copyright © 2017 Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx
OEE & TEEP let you keep score, but that’s about it.
• They’re useful, but reactive, not predictive
• What are the historical causes of poor OEE? Are they clearly
understood? Are they static or do things change over time?
• Are there ways to recognize patterns in historical data that can provide
advanced indication of those causes developing?
• Can you act on those causes?
• How much time would you have to act?
• What would you need to improve your ability to predict issues and act
on those predictions?
23Copyright © 2017 Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx
Discussion
Some potential improvements - beyond normal operations
Physical Web node functions
Vibration analytics for rotational and reciprocating machinery
Additional process quality instrumentation
Detailed / granular OEE data collection via SCADA and machine control integrations
Physical Web and Cloud Edge node functions
Event correlation analytics
24Copyright © 2017 Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx
This is the tip of the iceberg
A lot of critical questions have been left unanswered here. Great discussion topics!
Greenfield vs Brownfield (factory fit vs retrofit)
Remote device management
Compute capacity
System operations
Additional material:
McKinsey Analytics Report
http://www.mckinsey.com/business-functions/mckinsey-analytics/our-insights/the-age-of-analytics-competing-in-a-data-driven-world?cid=analytics-alt-mgi-mgi-oth-1612
Peter Levine – The End of Cloud Computing
http://a16z.com/2016/12/16/the-end-of-cloud-computing/
Frank Chen – Deep Learning and Machine Learning Primer
http://a16z.com/2016/06/10/ai-deep-learning-machines/
Thank You

Edge patterns in the IIoT

  • 1.
    EDGE PATTERNS INTHE IIOT BRAD NICHOLAS CHICAGO IOT MEETUP MARCH 2017
  • 2.
    2Copyright © 2017Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx AGENDA 01 3 minutes about Uptake 02 Some key considerations 03 The 3 patterns 04 Manufacturing discussion / Q&A
  • 3.
    3Copyright © 2017Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx Uptake at a glance AEROSPACE AGRICULTURE CONSTRUCTION ENERGY 104M predictions generated to date 2014 founded in Chicago 82% across Data Science & Engineering 700 Employees Uptake has developed partnerships in: HEALTHCARE MINING RAIL RETAIL Uptake selected as the hottest startup of 2015 – beating out Uber and Slack. – Dec 2015 Uptake’s Industry Thought Leaders featured in:
  • 4.
    4Copyright © 2017Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx Our platform is purpose-built to deliver actionable insights and recommendations into workflow, empowering people to create value Raw Data Data Ingestion Platform Apps Data Science Engines Data Integrity Software Development Kit Failure Prediction Anomaly Detection Recommendations Event / Alert Filtering Data Operations Center Normalization & Cleansing End to end visibility Encryption in transit and at rest API Portal Developer Content Mgmt. App Store Tools Assets Customers ERP Contextual • Weather • Social Media • 3rd party Sample Apps: • Condition- Based Monitoring • Supply Chain Optimization • Fuel and Energy Management • Performance Optimization Workflow Integration Examples: • Automated locomotive re-routing • Automated parts ordering • Automated maintenance scheduling END-TO-END CYBER, INFORMATION, AND OPERATIONAL SECURITY
  • 5.
    5Copyright © 2017Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx About Me I run the IoT team at Uptake bradn www.linkedin.com/in/bradn Automotive, Manufacturing, Consulting, Telecom, Startups EE MBA Fun fact: I “OEM+” hack & restore German cars
  • 6.
    6Copyright © 2017Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx We’re hiring. https://boards.greenhouse.io/uptake Come see me if you’re interested in IoT, device management, embedded programming, crypto
  • 7.
  • 8.
    8Copyright © 2017Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx Digitization is lagging in many industry sectors that need IIoT MGI Industry Digitization Index http://www.mckinsey.com/industries/high-tech/our-insights/digital-america-a-tale-of-the-haves-and-have-mores • Quasi-public and/or highly localized sectors are lagging in digitization • Labor-intensive sectors need digital tools for the workforce • Knowledge-intensive sectors are already highly digitized • Capital-intensive sectors have high IoT potential • Service sectors can digitize customer transactions • B2B sectors can benefit from expanded digital engagement 6 5 1 2 3 4
  • 9.
    9Copyright © 2017Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx 3 essential elements to IIoT value creation Data Ingestion “Sense” Analytics “Infer” Workflow “Act”
  • 10.
    10Copyright © 2017Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx IIC’s reference model for industrial analytics covers most of the bases Multi-tiered approach Sensing vs Actuating Different time horizons Open vs Closed loop Source: Industrial Internet Consortium IIRA http://www.iiconsortium.org/IIRA.htm
  • 11.
    11Copyright © 2017Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx Where you compute affects many things There is no one architecture that will address everything. But there are certainly some common questions to answer Proximity Response Time Node Computing Capacity Bandwidth Consumed Focal Points Exceptions Sense Act
  • 12.
    12Copyright © 2017Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx How you are able to connect also affects what you can do Latency, bandwidth, cost and complexity are usually not as optimal as you want them to be MobileLocal IndividualSite
  • 13.
    13Copyright © 2017Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx Other key IIoT needs, beyond strong security & viable economics Separation of Concerns is essential Key to managing complexity, achieving maintainability and resilience https://effectivesoftwaredesign.com/2012/02/05/separation-of-concerns/ IP protection is crucial Data rights management for both original and derived data, at rest and in flight, all nodes, including authorized use https://motherboard.vice.com/en_us/article/why-american-farmers-are- hacking-their-tractors-with-ukrainian-firmware Heterogeneity is unavoidable Computing environments Node state Mobility vs fixed location Networking options and node availability Domain responsibility IT/OT barrier is literally a real thing Operational control comes first Skills/expertise is very different Most capital equipment is decades old and relies on physical security http://blog.iiconsortium.org/2016/08/it-vs-ot-for-the-industrial-internet-two- sides-of-the-same-coin.html
  • 14.
  • 15.
    15Copyright © 2017Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx 3 patterns seem to address most IIoT deployment scenarios Physical Edge On-device IoT node Platform & Applications Cloud Edge reverse CDN for the physical web Edge Gateway “On location” connectivity node
  • 16.
    16Copyright © 2017Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx The Cloud Edge is effectively the ‘virtual physical web’ A hybrid node that serves as a “concentrator” or “reverse CDN” for the physical web. It can isolate IoT traffic and service cloud-based applications with anything they need from the physical web Concentrates physical web data streams Interacts with Edge Gateways and higher end Physical Edge nodes Serves web APIs to cloud applications You can train ML using the data on this node. You could continually train ML given sufficient compute capacity and data. You can distribute its contents via CDN, subject to data rights management
  • 17.
    17Copyright © 2017Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx The Physical Edge interacts directly with IIoT data sources Protects the OT layer and hosts specialized, “high interaction” IoT processes Serves as a direct data extraction point for physical web data generated by a machine or process Protects machine / process operation at all costs, even if data extraction compromised Runs on-machine / on-process analytics functions Protects OEM and machine owner IP by enforcing data rights management at the source, under terms suitable to the IP owners Must be designed and deployed in collaboration with machine / process OEMs and operators Provides much richer data access capabilities
  • 18.
    18Copyright © 2017Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx The Edge Gateway Resides in proximity to physical web nodes and handles connectivity gaps Manages “inter physical web” IoT interactions that aren’t needed to control things Primary function is to monitor physical web machines / processes Eliminates the need for physical web devices to interact with the Cloud Edge directly Queues on premise when backhaul connectivity is unavailable, restricted due to cost or otherwise unusable Speaks local machine dialects
  • 19.
    19Copyright © 2017Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx The 3 edge patterns can be implemented flexibly Physical Edge nodes deployed on advanced machines with excellent connectivity can connect directly to Cloud Edge nodes – without an Edge Gateway Cloud Edge nodes could be deployed anywhere connectivity to other edge nodes and “data center quality” bandwidth is available • A very high end physical web machine or process • At a fixed location like an airport terminal Edge Gateway nodes could be co-deployed with Physical Web nodes as long as suitable backhaul connectivity to a Cloud Edge node is available
  • 20.
    Discussion Applying the edgepatterns in manufacturing 04
  • 21.
    21Copyright © 2017Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx OEE - Overall Equipment Effectiveness Total Productive Maintenance Seiichi Nakajima 1982-1984 www.AMTonline.org http://capstonemetrics.com/files/whitepaper-oeeoverview.pdf OEE = Availability x Performance x Quality TEEP = Loading x OEE
  • 22.
    22Copyright © 2017Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx OEE & TEEP let you keep score, but that’s about it. • They’re useful, but reactive, not predictive • What are the historical causes of poor OEE? Are they clearly understood? Are they static or do things change over time? • Are there ways to recognize patterns in historical data that can provide advanced indication of those causes developing? • Can you act on those causes? • How much time would you have to act? • What would you need to improve your ability to predict issues and act on those predictions?
  • 23.
    23Copyright © 2017Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx Discussion Some potential improvements - beyond normal operations Physical Web node functions Vibration analytics for rotational and reciprocating machinery Additional process quality instrumentation Detailed / granular OEE data collection via SCADA and machine control integrations Physical Web and Cloud Edge node functions Event correlation analytics
  • 24.
    24Copyright © 2017Uptake23-Mar-17Brad Nicholas – Chicago IoT March 2017.pptx This is the tip of the iceberg A lot of critical questions have been left unanswered here. Great discussion topics! Greenfield vs Brownfield (factory fit vs retrofit) Remote device management Compute capacity System operations Additional material: McKinsey Analytics Report http://www.mckinsey.com/business-functions/mckinsey-analytics/our-insights/the-age-of-analytics-competing-in-a-data-driven-world?cid=analytics-alt-mgi-mgi-oth-1612 Peter Levine – The End of Cloud Computing http://a16z.com/2016/12/16/the-end-of-cloud-computing/ Frank Chen – Deep Learning and Machine Learning Primer http://a16z.com/2016/06/10/ai-deep-learning-machines/
  • 25.