IoT Reference Architecture and Deployment Alternatives
Juxtology Proprietary
End to End IoT Solutions
IoT Reference Architecture – Multiple AlternativesAllSeen, OneM2M, Industrial Internet Alliance, Open Internet Consortium, …
Juxtology Proprietary
The reference architecture converges on Devices, Gateways, Data Accumulation, Analytics and UI
World Forum’s more granular 7 Layer Model Microsoft Azure model is much more generalized
IoT “Standards” vary in terminology and focus – But, they provide good guidelines for platform deployment needs
Industrial Internet Consortium – Three tier architecture including Edge, Platform and Enterprise tiers
Open Connectivity Foundation (and iotivity.org) – 4 tiers: discovery , data transmission, device mgt, and data mgt
Open Interconnect, AllSeen – more HAN focused, less focus to analytics, cloud and business intelligence
OneM2M Standard – Three tier architecture including Network Services, Common Services and Application Layers
Due to its granularity the IoT World Forum Reference Model provides a good model to assess platform requirements
2
IoT Reference Model – Physical Devices and ControllersField Assets, Events, Objects – The “Things” in IoT
Juxtology Proprietary
This layer is a little ambiguous as legacy “Things” may need to be retrofitted with sensors and connectivity
Architectures and Implementation Alternatives:
Asset IoT solutions can be very simple or very complex
o Single sensor monitoring and reporting
o Multiple sensor monitoring, reporting and control
o Multiple sensor monitoring, reporting, intelligence and control
o Field based analytics, local intelligence, and fault management
Power / Cost / Range / Bandwidth all influence design decisions
Fast time-to-market solutions are typically loosely-coupled due to
engineering time / cost and regulatory / compliance issues
o Early deployments can be “multi-piece” solutions
o Market / design / P&L validation should trigger full integration
o Tiered solutions connects “capillaries” to GW to Cloud
o “Capillaries nodes” can connect directly to Cloud
o With integration, field equipment can direct connect to Cloud
Asset “Enterprise Cost” is a key motivator in IoT design
o “Cost” is enterprise cost (maintenance, asset cost, failure cost)
o High cost assets justify more IoT management investment
o Lower “cost” assets may force simpler, cheaper IoT solutions
o Critical assets may require very low latency response times
3
IoT Reference Model – Connectivity (Focus on Wireless)Communications / Connectivity can be Embedded or Wired to Assets
Juxtology Proprietary
Multiple technologies, design alternatives and suppliers makes for a rich but complex ecosystem
Important Connectivity Considerations:
Power is critical to many applications, and is a
range/bandwidth/availability trade-off
Multi-tier solutions can be more complex, but can
optimize cost/range/availability issues
Mesh technologies are also more complex but provide
higher availability
Range / Power / Bandwidth Trade-offs:
LTE for high BW distributed assets (mode 0 coming)
LoRa / SigFox for low bandwidth / low power / long range
WiFi high power / low range / high speed
WiFi 802.11ac coming with power & range benefits
802.15.4g mesh is a good range / power / BW alternative
Very short range: Bluetooth, WirelessHART and ISA100
4
IoT Reference Model – Connectivity (Focus on Wireless)
Communications / Connectivity can be Embedded or Wired to Assets
Juxtology Proprietary
The nature of the IoT requirements will drive Field Area Network (FAN) architecture
Bandwidth
Ra
ng
eSingle or Multiple Connectivity Options may be Used:
Where practical (power & coverage) LTE is simple
Power / Serviceability may force use of LPWAN
Bandwidth needs may force multiple technologies
Hierarchical field area networks are a design pattern:
oLow cost / low power “Edge Nodes” connect to GW
oGateways provide “Edge Intelligence” and backhaul
Range / Power / Bandwidth Trade-offs:
LTE for high BW distributed assets (mode 0 coming)
LoRa / SigFox for low bandwidth / low power / long range
WiFi high power / low range / high speed
WiFi 802.11ac coming with power & range benefits
802.15.4g mesh is a good range / power / BW alternative
Very short range: Bluetooth, WirelessHART and ISA100
5
IoT Reference Model – Connectivity (Focus on Wireless)
Communications / Connectivity can be Embedded or Wired to Assets
Juxtology Proprietary
The nature of the IoT requirements will drive Field Area Network (FAN) architecture
Backhaul Network
(Hybrid)
LTE
Licensed or
Unlicensed Private
Long Reach Low Access
(LoRa, Mesh, Others)
Wired
Alternative Deployment Scenarios
Technologies SigFox LoRa LTECat-M 802.11ah Dash71.0 IngenuRPMA nWave Zigbee
Frequency868/902ISM
433/868/780/91
5ISMCellular
sub-1GHz
unlicensed
433/868/915
ISM2.4GHzISM sub-GHzISM
2.4GHz,915,
868ISM
ChannelWidth UNB 8-64/125kHz 1.4MHz 1-16MHz 25-200KHz 1MHz UNB
Range30-50km
2-5Kmurban
15Krural2.5-5km 1km to5km 500kmLoS 10km 300m
EndNodePower
10uW-100mW
dBm<27dBm 100mW 1mW-1W TBD to20dBm 25-100mW 0dBm
Uplink
100bpsto140
msg/dayto100kbps 200kbps 150-350kbps 9.6to166kbps 624kbps/sector 100bps 250kbps
Downlink4msgof8b/day to100kbps 200kbps 150-350kbps 9.6to166kbps 156kbps/sector -- 250kbps
Devices/AP 1M 100K 20k 8191 connectionless 384k/sector 1M 65k
TopologyStar staronstar Star Star,Tree
Node-Node,
Star,Tree
Star,treew/
extenderStar Mesh,Star,Tree
Roaming Yes Yes Yes w/802.11r Yes Yes Yes
Status Deployment Deployment 2016 2016 Deployment Deployment Deployment Deployment
Comparison of IoT Radio Technologies
Cloud Application
Tiered
Solution
Direct
ConnectFully
Integrated
Gateways
Edge Nodes
Sensors
Edge Nodes
Sensors Embedded
6
IoT Reference Model – Edge ComputingEdge Computing may be at the “Edge Node”, Gateway, or Cloud Edge
Juxtology Proprietary
Intelligence at the edge enables bandwidth limiting, local autonomy, and low latency feedback loops
Edge Computing Architecture and Design Considerations:
Power availability & serviceability for the IoT deployment
Real-time feedback, analysis and control needs
Requirements for intelligence at the edge:
o Local data processing to minimize connectivity bandwidth
o Local decision making to address real-time control needs
o Customer configurable or designed edge applications
Device “Ruggedness” requirements (thermal, shock, EMI, etc.)
Over-the-air updates and bandwidth required for updates
Overall Gateway / Edge Node / purpose suitability:
o Flexibility to handle multiple sensor input types
o Low-latency detection of trigger events in “moving data”
o Cost effectiveness for purpose (edge vs GW vs computing)
Design and Architectural Drivers:
OS – Linux or small footprint embedded OS
CPU – ultra-low, low, or higher power (data pipe or analysis)
Connectivity – Range + BW = higher power needs
These considerations often drive the need for multi-tier edge node
plus gateway (edge processor) architectures as mentioned before.
Cloud Application
Tiered
Solution
Direct
ConnectFully
Integrated
Gateways
Edge Nodes
Sensors
Edge Nodes
Sensors Embedded
7
IoT Reference Model – Edge ComputingEdge Computing may be at the “Edge Node”, Gateway, or Cloud Edge
Juxtology Proprietary
Intelligence at the edge enables bandwidth limiting, local autonomy, and low latency feedback loops
Operating Systems
IoT and Gateway Hardware
Distributed Intelligence & Control
Connectivity
Protocols & Translations CoAP
Wired ISP
Sensors and I/O Connectivity
Fie
ld I
oT
So
luti
on
Co
mp
on
en
ts
Field Assets, Smart Cities, Smart Grid
Microwave
There is a Sea of Options for Edge Computing:
The needs of the IoT solution MAY drive very
specific set of technologies – but MAY NOT
Many solutions can leverage simple off-the-
shelf solutions from OEM suppliers
For more demanding solutions a design can be
created from specific ecosystem technologies
8
IoT Reference Model – Data AccumulationManaging the high Velocity, Volume and Variety of IoT Data
Juxtology Proprietary
This layer must handle not only data accumulation, but also real-time data distribution to applications
The ”Data Accumulation” layer is critical to IoT Systems and for
more complex systems may be significantly oversimplified:
Per the reference architecture, this layer is for data storage
But this “Cloud Edge” layer may add other needed functions in
beyond just storing field data in a “data lake” for later processing:
o DDS middleware such as publisher / subscriber, data
classification, filtering and in-motion transformation capabilities
o Low latency data screening, aggregation, and transformation
for critical system / low latency operations
o Data parsing and mapping to multiple DB types and schema
o Pre-processing between IoT and legacy application syntax
Some IoT solutions will require only simple SQL management
High Volume, Velocity, Variety IoT solutions will require NoSQL
Highly scalable solutions will require DDS and device intelligence
Architectural Requirements for Data Accumulation Layer:
High availability and reliability suitable to the application
Performance matched to velocity, volume and latency needs
DB ETL / Schema management to simplify upstream data mgt
Efficient services tied to data abstraction, applications and I/O
System design that enables ingress/egress latency needs
Data Accumulation / Storage Key Issues:
IoT data is Big Data with Big Data problems of
Volume, Velocity and Variety
High volume, velocity, variety assessment of in
motion data is an important IoT design aspect
Need Hot path - immediate processing & Cold
path for storage and application processing
9
Fie
ld A
sse
t D
ata
IoT Reference Model – Data AccumulationManaging the high Velocity, Volume and Variety of IoT Data
Juxtology Proprietary
Unique requirements (3V’s) will drive the simplicity or complexity of storage and tool complexity
Data Accumulation / Storage Key Issues:
IoT data is Big Data with Big Data problems of
Volume, Velocity and Variety
When IoT solutions are integrated into legacy
systems interoperability is important
High volume, velocity, variety assessment of in
motion data is an important IoT design aspect
Compute Clusters
Infrastructure …aaS Big Data Storage …aaS
Big Data Access and Control
Unique
Applications
Platform …aaS
The Data Accumulation Architecture Drives Decisions:
”3V” IoT drives need for Cloud-based NoSQL or hybrid solutions
In-memory high speed processing provides low-latency responses
Small enterprise (some Utility) may require on-premise solutions
Winning OEMs / SPs will provide solutions that provide:
Proven “canned” solutions that are easy to create and deploy
Tools to assist customers in designing data storage design
Data Distribution Services to manage disparate system needs
10
IoT Reference Model – Data AbstractionThe Data Abstraction Layer ”Makes Sense of the Data” for Applications
Juxtology Proprietary
Data reduction, context creation, augmentation with meta data, derived data creation …
The data density of IoT systems (Velocity, Volume and Variety)
makes data abstraction and overall volume reduction essential
Data processing in the Data Abstraction Layer transforms the high
volumes of data from IoT readings into more valuable information
For example, IoT readings are transformed from multiple
readings to levels, trends, conditions, exceptions and classified:
o Temperature readings reduced to “instance counts”
o Time series events reduced to relevant change vectors
oDevice data is augmented with meaningful IoT meta data
oAnomalous or critical readings can be queued to fast paths
This process is executed through the cleanup and processing of IoT
ingress data and creating the abstract data formats for applications
Multiple micro-services can be used for data transformation and
as front-end logic to application services
Appropriate ETL or Big Data operations transform raw data-lake
messages into formats and queues for applications
Multiple data abstractions may be created for the different types
of back-end applications (SQL / NoSQL) and latency needs
Virtual device and virtual device state models are maintained
based on the abstracted data and accessible to applications
Raw Sensor Data
Aggregation and Filtering
Metadata Integration
Abstraction and Reduction
Storage of IoT Models
11
IoT Reference Model – Application LayerReporting, Analytics and Control – and Integration into Back End Systems
Juxtology Proprietary
Initial IoT Application Layer may focus on “Things” but integration into the larger business processes is the goal
The core value of IoT is connectivity to business logic and
analytics. Get the data, manage it, ANALYZE it - then ACT!
The primary business drivers for IoT fall into three main categories:
Improved customer awareness, accessibility and engagement
Improved operational efficiency, profitability, and predictability
Advancement of social good with technology delivered services
The Application Layer Software Services deliver this value:
Business Logic Applications from consumer transactions to real-
time industrial controls deliver new levels of value with IoT data
Analytics applications identify trends from consumer behavior to
field asset operation and enable new business insights
Improved “resolution” from real-time IoT data allows rules based
engines to deliver higher value control, prediction and insight
Machine learning algorithms identify and exploit trends and
correlations previously hidden in the high volume (3V) IoT data
Automated report generation enhances compliance requirements
IoT is processed and transformed and fed in to advanced field
asset, factory equipment, fleet, or logistical control software
Data Visualization, Reporting, Analytics provide insight and
connectivity across workers – from the office to the field
12
IoT Reference Model – Application LayerReporting, Analytics and Control – and Integration into Back End Systems
Juxtology Proprietary
Combine near-real-time IoT information with historical operations data for next gen analytics
Real Time IoT Advances will Require Edge Intelligence, Low
Latency, and Fast-Path Applications in Concert “Real-time awareness” and event- based systems create new
opportunities for businesses deploying end-to-end IoT systems
Asset location, monitoring, specialized diagnostics , real-time telemetry,
usage pattern data become available for applications and analytics
Next gen tools that leverage real-time data put new power in machine
data analytics, operations personnel and business users
Fusing real-time data and historical structured and unstructured data
creates a new dimension of operational and real-time intelligence
Real-time, state-based visibility will transform automation and efficiency
with closed-loop rules based systems (on-demand or event-driven)
Machine intelligence applications will eliminate run to failure operations
and deliver less human attention and more economic value
Rea
l Tim
e A
cce
ss
13
Source: Vitria.com
IoT Reference Model – Application LayerReporting, Analytics and Control – and Integration into Back End Systems
Juxtology Proprietary
Move from monitor, to control, to system level optimization and to transformative business insight
The full value of IoT integration will evolve over time
Initial focus is simple connectivity, data acquisition, condition
monitoring and control – moving to performance optimization
Automation and closed loop control is the next wave
Building IoT and near-real-time results into business operations
applications will bring the biggest wave of economic value
Architecture
IoT and old structured Data
Source: GE Automation
14
IoT Reference Model – Application LayerReporting, Analytics and Control – and Integration into Back End Systems
Juxtology Proprietary
IoT Analytics has become an industry to itself with an entire ecosystem of tools, suppliers and consultants
15
IoT Reference Model – UI, Collaboration, and ProcessUsing IoT Information / Visualization / Analytics to Transform Business Operations
Juxtology Proprietary
From simple operational visualization, to advanced IoT control systems, to new process creation
Connect People with Data Portals and Business Processes -
Integration of IoT information with other business logic and
Application Enablement Platforms and Total System Visualization is
where the value of IoT deployments is realized and enhanced
Key Operations metrics, Asset information, Consumer behavior
become tangible with near-real-time visualization
Interactive visualization tools enable hands-on inspection and
analysis to uncover issues and discover event correlations
IoT device policy, rules and state management tools enable
optimization of operations and system lifecycles
Application Enablement Platforms simplify integration of new
data and derived data to enhance back end business applications
Field maintenance actions and prioritization based on near-real-
time information transforms the efficiency of field personnel
Architectural and deployment considerations:
RESTful web services to orchestrate IoT information with
enterprise applications and cloud services
Automation enablement and exception based alarms through
rules engines using IoT and business application integration
Customize and optimize interfaces for specific employee needs
Create workflows based on near-real-time events / transactions
16
IoT Reference Model – UI, Collaboration, and ProcessUsing IoT Information / Visualization / Analytics to Transform Business Operations
Juxtology Proprietary
From simple operational visualization, to advanced IoT control systems, to new process creation
Combine historical data and real-time IoT data to
create new system modeling and business insight
ThingWorx
17
Bo
sh
Inte
l -
IoT
Te
xa
s In
str
um
en
ts
AR
M
Sm
art
Worx
Lib
elli
um
Ma
ch
Fu
Me
sh
ify
De
vic
e W
ise (
Te
lit)
AT
&T
–M
2X
Ve
rizo
n
Vo
da
fone
Ae
ris
IBM
–W
ats
on I
oT
Mic
roso
ft A
zu
re Io
T
Am
azo
n Io
T
Go
ogle
Io
T
Inte
rDig
ita
l
Vitria
Ge
ne
ral E
lectr
ic
Sie
me
ns
AB
B
Sa
lesF
orc
e.c
om
PT
C / T
hin
gW
orx
Presentation / Visualization
IoT Reference Model Coverage – Few Suppliers Provide “Whole Products”Customers must work with multiple IIoT solution providers to deploy solutions
This graph provides a subjective view of IoT function coverage based on recent Juxtology research
End-User Integration Services
IoT Solution Designer Software Tools
IoT Cloud-connect / Dev. Mgt.
IoT LTE Transport
IoT Gateway / Aggregation
Field Area Networks
Edge Intelligence
IoT Edge Devices
Sensor Integration
Juxtology Proprietary18
Processing / Analytics
Machine Learning
Strong Differentiation
Emerging Leaders
Strong Players
Solutions for IoT Remain Very FragmentedCustomers still must work too hard to field working solutions
Solution providers that make IoT easy for customers will have critical competitive advantage
Compute Clusters
Presentation
Infrastructure …aaS Big Data Storage …aaS
Big Data Access and Control
Cloud-based
Applications
IoT Analysis and Insight
Web Interface and Device Control
Operating Systems
IoT and Gateway Hardware
Distributed Intelligence & Control
Connectivity
Protocols & Translations CoAP
Wired ISP
Sensors and I/O Connectivity
Clo
ud
Io
T S
olu
tio
n C
om
po
nen
ts
Fie
ld I
oT
So
luti
on
Co
mp
on
en
ts
Platform …aaS
Transport: Wired and Wireless Private Networks
Field Assets, Smart Cities, Smart Grid
Users and
Customers
Microwave
Juxtology Proprietary19
End
Juxtology Proprietary