Chapter 1

What Is loT?

1 can think of is online and communicating

about anything you
lma0ine a world where just order to enable new services that enhance our lives. From
other things
and people in clothing monitoring
to order to sensors in your
drones delivering your grocery shift forward.
self-driving
know is set to undergo a major technological
the world you
vour health,
as the Internet of Things (loT).
known collectively
This shift is This means that
the unconnected."
to "connect
and goal of loT is will
The basic premise
to a computer network, namely the Internet,
currently joined
objects that are not
and interact with people and other objects.
connected so that they can communicate us tosenseand controlthe
be in which deviceswillallow
transition intelligent
o lOTis a technology smarter and connecting them through an
physical world by máking
obiects
network'
across a network,
a
controlled remotely
machines can be sensed and allows for
When objects and is enabled. This
the physical world and computers of
tighter integration between
enablement
automation, and the
of efficiency, accuracy,
improvements in the areas

0 advanced applications. somewhat

even find it
multifaceted, and you may
The world of IoT is broad and that it encom-
and
of components protocols
complicated at first due to the plethora domain, it is good
to view it as
loT as a single
technology times
of which are at
passes. Instead of 'viewing and technologies, all
elements is
or various concepts, protocols,
Ca While the wide array of loT
particular industry. automation, at the
Omewhat dependent on a the of productivity and
numerous benefits in
areas
numbers of
devices and
create
as scaling
the vast
new challenges,
such
tof ntroduces
amounts data that need to be processed.
 loTImpact
Projections on the potential impact of loT are impressive. About 14 billion, or
o f "things" are connected to the Internet
just0.06%
today. Cisco Systems predicts that by 2020, tnis
3 number will reach 50 billion. A UK government report speculates that this number coud
estimates that
CvEn higher, in the range of 100 billion objects connected. Cisco further
 Is loT?
8 Chapter 1: What

trilnon inprofits and cost savings Figure 1-2

connections will lead to SI9
these new in the number of devices being connected.
look at the growth
nrovides a graphical
fundamentally shift the way people
and
mean is that lol will
What these numbers surroundings. Managing
and monitoring smart objects
with their
businesses interact whole new level of
data-driven decision making.
enables a
using real-time
connectivity and delivers new
optimization of systems and processes
in the
This in turn results and businesses while improving
the overall
that savetime for both people
services
quality oflife. and their impact. These
some of the benefits of loT
Thefollowing examples illustrate view of practical loT use cases to clearly
with a high-level
examples will provide you life. For more in-depth
use cases, please
refer to
illustrate how loT will
affect everyday
"loT in Industry."
the chapters in Part II,

50 "Thnp "per person iae0a03 50

Billion
Objects
40
st.cture
51 faste than electricity &telephamy

30

20
Inflection Point

World Population
10
6.307 7.83
7.347
6721 6 894

2020
2003 2008 2010 2015
net
FIgure 1-2 Tbe Rapid Growth in the Number of Devices Connected to the mtc
 loT Irnpact 9

Google

Figure 1-3 Google's Self-Driving Car

Basic sensors reside in cars already. They monitor o1l pressure, tire pressure, temperature,
and other operating conditions, and provide data around the core car functions. From
behind the steering wheel, the driver can access this data while also controlling the car
using equipment such as a steering wheel, pedals, and so on. The need for all this sensory
information and control is obvious. The driver must be able to understand, handle, and
make critical decisions while concentrating on driving safely. The Internet of Things is
replicating this concept on a much larger scale.

Today, we are seeing automobiles produced with thousands of sensors, to measure every
thing from fuel consumption to location to the entertainment your family is watching
during the ride,. As automobile manufacturers strive to reinvent the drivingexperience
these sensors are becoming IP-enabled to allow easy communication with other systems
both inside and outside the car. In addition, new sensors and communication technolo-
gies are being developed to allow vehicles to "talk" to other vehicles, traffic signals,
School zones, and other elements of the transportation
ing to
infrastructure. We are now start
realize a truly connected
transportation solution.
MOst connected roadways solutions focus on resolving today's transportation challenges.
nese challenges can be classified into.the three categories highlighted in Table 1-2.
Jable 1-2 Current
Challen, Being Addressed by Connected Roaduays
Challenge Supporting Data
Safety
NCCording to the US Department of Transportation, 5.6 million crashes
WEre reported in 2012 alone, resulting in more than 33,000 fatalities.
01 and the enablement of connected vehicle technologies will empower
urivers with the tools they need to anticipate
potential crashes and
STgnificantly reduce the number of lives lost each
year.
10 Chapter 1: What ls loT?

Supporting Data
Challenge vehicle
More than a billion the roads worldwide. Connected
cars are on
Mobiliry operators and drivers
to make
mobility applications can enable system
more informed decisions,
which can, in turn, reduce travel delays.
of travel delay per year, and reducing
Congestion causes 5.5 billion hours
communication
travel delays is more critical than ever before. In addition,
between mass transit, emergency response vehicles, and traffic management
infrastructures help optimize the, routing of vehicles, further reducing
potential delays.
Environment According to the American Public Transportation Association, each year
transit systems can collectively reduce carbon dioxide (CO,) emissions
by 16.2 million metric tons by reducing private vehicle miles. Connected
vehicle environmental applications will give all travelers the real-time
information they need to make "green" transportation choices.
Sources: Traffic Safety Facts, 2010; National Highway Traffic Safety Administration, June 2012; and
WHO Global Status Report on Road Safety, 2013.

By addressing the challenges in Table 1-2, connected roadways will bring many benefits
to society. These benefits include reduced
traffic jams and urbancongestion, decreased
casualties and fatalities, increased response time for emergency vehicles, and reduced
vehicle emissions.
Forexample, with IoT-connected roadways, a concept known as Intersection
Assist Movement
(IMA) is possible. This application warns a driver (or
response in a self-driving car) when it is not safe to
triggers the appropriate
enter an intersection
due to a high
probability of a collision-perhaps because another car has run a
into the wrong lane. Thanks to the stop sign or strayed
communications system between the vehicles and the
infrastructure, this sort of scenario can be handled quickly and
a
graphical representation of IMA. safely. See Figure 1-4 for

Intersection Movement Assist

warns drivers if it is not safe to
enter an intersection.

iiiin
Figure 1-4 Application of Inrersection Movement Assist
 loT Challerigos 23

oTChallenges
eles
paints an impressive
cture, it docs not come without SIg
abled future
W h i l ea n l o 7 - e n a b l e d

become reality, but certain obstacles need to

of loT have
Aany parts
nificant challenges. Man roughoutindu
throughout industry and our everyday life.
ubiguitous
to become
Overcome
for lol
of the most s significant challenges and problems that loT is
he a few
4 highlights
unentlyfacing

Challenges
loT
1-4
able Description

ofhérwork3
can be scale of OT can be several
large, the

While the scale

Challenge

example, one largeclectricalutilityin Asia

Oorders of magnitude
larged For
Scale

deploying Pv6-based smart

metets on its(electrical grid.
recently began of thóusandsof employees (which
company has tens of meters in the
While thisitility in the network),.thenumber
considered IP nodes the
can be
his thescal of the network
means
ofmiliens
service area tens Chapter 5,
is
more than 1,000-fold!
hasncreásed by
(utility18 managing approaches are
how new design
Network Layer, explores
Pas the loT to IPvó networks
into the millions of devices.
being developed
scale
"things" and people,
becoming connected with other surface is now
With more "things" issue for loT. Yourthreat
Security (ncreasingly comple amajor
security is an
device getshacked, its connectivity is
greatly expanded,
andfa a)aunching pointtoattack
device can[erve as about
concern A compromised pervasive across just
security is also
other devicesand
systems.joT loT security, seè Chapter
8,
more information on
facet of loT/For
every
"Securing loT") of the data
everydaylives, much
our

As sensors becomemore prolifjgin theitactivities. This

data
Privacy toindividualsand
the gatherwillbespecific
health information
to shoppingg patterns and
transactions
canrange from this data hasmonetaryvalue.
For businesses, individuals
at a fetail establishmentdiscussing who owns this data and how
now
Organizations are
andvith whom)
can control whether itigShared
iségoing totrigger deluge
a of data that,
Big data and lol and its large number of sensors
and insight
willprovide)critica(information
This data
data analytics must be handled. however,is
in an efficient mannen The challenge,
be processed in
t itcan dataarrivingJrom different sources
évaluatingmassive amounts of more 7 for
timelymanner See Chapter
doing so in a
Naious forms and data perspective.
itfacés from a big
nformationon loT and the challenges
24 Chapter What
1: Is loT? rop fas

Description
Challenge
As withany other
nascent technology{ various protocols and
nd architectures
Interoperability
are ockeying for
market share ana standardizationwithin lotc
basee-o propretaryelem ofF
o
these protocols and architectures are
thisments,
others are open. Recentlol standarare helpingiminimizethis nrLd and
problem,)
butthere are pften variousprotocOlsand implementations availa or
oT nerworksThe prominent protocols and architectures-especialy
open, standards-based implementations-are the subject of this booi
For more information on loT architectures, see Chapter 2, "IoT Netwo.
rk
Architecture and Design." Chapter4, "Connecting Smart Objects."
Chapter 5, TP as the loT Network Layer," and Chapter 6, "Application
Protocols for IoT," take amore in-depth look at the protocols that make up
loT
T

Summary
This chapter provides an introductory look at the Internet of Things and answers the
question "What is loT?" loT is about connecting the unconnected, enabling smart
to communicate with other objects, systems, and people. The end result is an objects
network that allows more control of the physical world and the intelligent
enablement of advanced
applications.
This chapter also provides a historical look at loT, along with a current view of loT
as the
next
evolutionary phase of the Internet. This chapter details a few high-level use cases
show the impact of loT and some of to
the ways it will be changing our world.
A number of loT
concepts and terms are defined throughout this chapter. The differ-
encesbetween loT and
and OT. The last sectiondigitization discussed, as well as the convergence between IT
are
details the challenges faced by loT.
This chapter
should leave you with a clearer
addition, this chapter serves as
the
understanding of what loT is all about. In
into loT in the foundational block from which you can dive
following chapters. further
 boesr- o ank
Comparing loT Architectures 35

and end-
Network.layer ís is the communication domain for the loT devices
pointsItincludes the devices themselves and thecommunications networy that links
them.Embodimentsof this communicationsinfrastructureinclude wirelessmesh
such
technologies, such as IEEE 802.15.4, and wireless goint-to-multipoint system3)
as IEEE 801.1 1ah. Also included are
wired device connections, such as IEEE 1901

power line communications. Chapter 4 provides more details on these connectivity

technologies.
devices communicate with
In many cases, the smart (and sometimes not-so-smart)
communication is not necessary, and the
eachother. In other cases, machine-to-machine
devicessimplycoimunicate througha field area network (FAN to u[e-case-specíffc

apps i thdloT applicationdomain. JTherefore, the device domain also includesthe

into the core network and acts
gareway device, which provides communications up
as a demarcation point between the device and network domains

by oneM2M covering loT

Technícal Specifications and Technical Reports published
at www.onem2m.org.
found
functional architecture and other aspects can be

Architecture
TheloT World Forum (loTWF) Standardized
Automation,
I n 2014 the loTWEarchitectural committee (led by Cisco,TBM, Rockwell various IoT
model While
and others) ublisheda seven-layefloT architecturalreterence
6ffersa clean, simpli-
referencemodels exist,theone put forth by the loT World Forum and access. It pro-
f i e d perspective on loT and includes edge computing dataatorage,
Each of the.seven
vides asuccinct way ofvisualizing loTfrom atechnicabperspective.
encompasses theentire model)
layersis broken down into specific functions, and(security
Model published by the loTWF. A
Figure 2-2 details the loT Reference

Levels

Collaboration & Processes

(Invoving People & Business Processes) Center
Application
(Reporting, Analytics, Control)

Data Abstraction
(Aggregation & Access)

Data Accumulation
(Storage)

Edge Computing
(Data Element Analysis & Transfomation)

Connectivity
(Communication & Processing Units)

Physical Devlces &Controllers

Edge
Sensors, Devices, Machines,
(The Things" in loT) Inteiligenl Edge Nodes of aii lypes

Figure 2-2 loT Reference Model Published by the loT World Forum
 and
36 Chapter 2: loT Network Architecture
defines a
set f levels with
of
Model dedicata
Or a
lol
Refcrence
coud service dl data
As shown in Figure 2-2, the either a other tu
could be machines,
and
ypes.
(this
fro the
trol flowing trom center devices,
sensors,
originating from
th
incudes
the stack,
center), to the edge, which
general,
up are able tto
data travels r e f e r c n c e modcl, we are
able
achieve
In
intelligent end nodes.
center.|Using
this
to the
and goes orthbound

the following:
parts
into smaller
thedOT problen) eto one
one another
Decompose
technologiesat
each layer
and how they relate)to
(ldentifydifferen
be provided by different vendors
differenpartscan
which
D Define a systemjn to interoperability
defining(interface_ thae leads
Have aprocess pf transition points between leve
at the
Define atiered security model that isenforced,
of the loOT
closely at each of the s e v e n layers
sections look more
h e following
Reference Model.

1: Physical Devices
and Controllers Layer
Layer Reference Model is the physical devices
and eóntrollers
The first layer of the loT of Things including the various
home to the "things"in theIprernet
layer.This layer is jnfoURmation. Thesize of
and receive
these
endpoint devices and that(send
sensors
in a factory. Their
can range from almost microscopicsensors togianthnachines ànd/or
"Things and beingcapable of being
queried controlledh
primary function isgenerating data
over a network.

ot
Layer 2: Connectivity Layer
Inthesecond layer of the IoT Reference Model, thelfocus is on connectivity. The most

importan functionpf thisloT layer is the reliable and timel transmission of data.
More specitically.chis includeytransmissions between Layer 1 devices and the netwórk
and between the netwokandinformation processingthat occurs at Layer 3 (the edge
computinglayer).
As you may notice, the connectivity layer encompasses all
networking elementsof
loT and doesn't really distinguish between the last-mile network (the network between
the sensor/endpoint and the loT gateway, discussed-later.in this chapter), gateway, and
backhaul networks. Functions of the connectivity layer are detailed in
Figure 2-3.
 Comparing loT Architectures 37

Connectivity
(Communication and Processing Units)

Layer 2 Functions:
Communications Between Layer 1 Devices
Reliable Delivery of Intormation Across the Network
Switching and Routing
Translation Between Protocols
Network Level Security

Figure 2-3 loT Reference Model Connectivity Layer Functions

Layer 3: Edge Computing

Layer
LEdge computing is the role of Layer 3, Edge computing is often
layer and is discussed in the section referred to as the "fog"
"Fog
theemphasis is ondará reduction and Computing laterdatain thís chapter. At this layer
that is ready for
storage and procéssing convertingnetwork
byhigher layers.
flowsintoinformation
One of thebasic principlesyof
this peference model is that information
edge of the networkas possible Figure proçessingisínitiatedjas
24 highlights
earty and as close to the
ofthe loT Reference Model. the functions handled by Layer 3

Edge (Fog) Computing

(Data Element Analysis and Transformation)

Layer 3 Functions:
Evaluate
Data
and Reformat
for Processing at
Higher Levels Data Ready for
.Filter Data to Reduce Processing at
Traffic Higher Level Higher Levels
Processing
Assess Data for
Alerting,
Notification, or Other Actions

Data Packets
Figure 2-4 IoT Reference Model Layer 3 Functions
 38 Chapter 2: loT Network Architecture and Design

that occurs at Layer 3 is th evaluation of data to see if it

Another important function
sent to a higher layer. his also allows for
can
betiltered or aggregate) before being by other systems easie
to be reformatted or decoded.(making.additional processing
to see if predetined thresholds are cra
Thus, a critical function is assessing the data
andanyaction or alertsneed to be sent.

Upper Layers: Layers 4-7

IoT data generated by the bot
LThe upper layers deal with(handling and processing)the
tom layer. For the sake of completeness, Layers 4-7 of the IoT Reference Model are

summarized in Table 2-2..

Table 2-2 Summary of Layers 4-7 of the loTWF Reference Model

loT Reference Model Layer Functions

Layer 4: Data accumulation Captures data and stores it so it is usable þyappliçations

layer when necessary. Converts event-basedyfata tdjqueryabase
processing.
Layer 5: Data abstraction layer Reconciles multiple data formats and ensures consistent
semantics from varioHssources Confirms that the data
set is completg.ancconsolidate) data into e plac or
multip/é dat storesusing virtualization
Layer 6: Applications layer Interprets data)isingsoftwar applications. Applications
r, Control, and provide reportsbased onthe
analysis of the data.
Layer 7: Collaboration and (Consumes and share theapplication information.
processes layer Collaborating on and communieating-toT-information oten
requiresmultiplesteps and it is what makes loTusetu
S ayer can(hange business processey and delivers the
benefitspf loT

IT and OT Responsibilities in the loT Reference Model

An interesting aspect of visualizing an loT
architectlure this way is that you can start
organize responsibilities along IT and OT lines. Figure 2-5 illustrates
a natural demarca
tion point between IT and OT in the loT Reference
Model framework.

C o i ' r
 The Core loT Functional Stack 43

thisbook isthat loT isnot just about the control of loT devices but, rather, the(useful
insights gained from the datayenerated bythose devices Thus the applications layer
typically has both analytics an industry-specifi loT controlsystemcomponents.
You will notice that security is central to the entire architecture, both from network con-
nectivity and data management perspectives. The chapters in Part I, "Engineering loT
Networks," discuss security at cach laycer. Chapter 8 is dedicated to the subject of securing
loT systems. The industry chapters in Part 11, "loT in Industry," highlight how lessons
learned in Parts 1, "Introduction to loT" and Il can be applied to specific industries. Each
of the Part l chapters examines the issue of loT security for a particular sector.
The architectural framework presented in Figure 2-7 reflects the flow of the chapters in
this book. To help navigate your way through this book, chapter numbers are híghlighted
next to the various layers of the stack.

The remainder of this chapter provides a high-level examination of each layer of this
model and lays the foundation for a detailed examinatión of the technologies involved
at each layer presented in Part I, and it gives you the tools you need to understand how
these technologies are applied in key industries in Part II

The Core loT Functional Stack

LOTnetworks ardlbuilt aroundhe conceptof "things," or smart objects(aerforming
functions and (deliveringhew connectedservices.These objects are "smart"becausethey
use a combination of contextualinformationjand configuredgoalsyo perform actions.
Thése actions can be self-contained(that is, the smart object does not rely on external
systems for its actions); however, in most cases, the "thing" interacts with an external sys-
tem to report information that the smart object collects, to exchange with other objects,
or to interact with a management platform. In this case, the management platform can be
used to process data collected from the smart object and also guide the behavior of the
smart object| From an architecturalstandpoint,Geveralcomponentshave to work together
for an laT network to bé operational:

Things" layer: At this layer, the physical devices need to fit the constraints of the
environmentmwhich they are deployed while still being able to provide theinforma
tion needed.
Communications network layer When smart objects are not self-contained, they
need to.communicatewith an external system, In many cases, this communication
uses a.wireless technology. This layer has four sublayers:
(Accesshetworksublayer: The last mile of the loT nerwork is the access network
Tis.is.typically.madeup of wireless technologies such as 802.11áh, 802.15.4g,
and LoRa. The sensors connected to the access network may also be wred.
 Design
A r c h i t e c t u r e
and
44 Chapter 2: loT Network
sublayer:
A
common

nunication system
communication

and
backhaul
network
in a given
area
around a common e a t
gateway.T
Gateways
s m a r t objects
The role of the
role of t
s m a r t objects.
organizes multiple with the medium iro
y
gateway
directly
c o m m u n i c a t e s

information
longer-range
through a
is processed
ed the
to forward the collected
central station
where the
information

which
This
the reae.
is the
is
headend function, reason this
backhaul) to a (application)

exchange is a Layer 7 also torwards parb

vackets
intormation
networks,
this gateway
gateway
On IP acts as a
router.
is called a it therefore
object another, and
from one IP network to networ.
to be successful, and
communication
sublayer: For. implemented to sunn.
support
Networktransport IP and UDP must be
protocols such as
layer
transport to use.:
and media
to connect
the variety of devices
Additional protocols must be in
place to
lIoT network management
sublayer:
to exchange
data with the sensors. Examples
allow the headend applications
include CoAP and MQTT.
needs to process
and analyticalayer: At the upper layer, an application
Application but to
control the smart objects when necessary,
thecollecteddata, noronly to on the information collected and, in turn, instruct
make intelligentdecision based
the analyzed conditions and change their
the "things" or other systems to adapt to
behaviors or parameters.

The following sections examine these elements

and help you architect your IoT commu,
nication network.

Layer 1:Things: Sensors and Actuators Layer

Most IoT networks start from the object, or "thing," that needs to be connected.
Chapter 3.Smart Objects: The Things in IoT" provides more in-depthinformation
about smart objects.
Erom an architecturalstandpoint, the variety of smart object tyP
shapes, and needsdrive the variety oflaT protocols.and architectures! There are my riad

ways to classify smart objects 9nefarchitectural classification çould be:

Battery-poweped or power-connected:This classification is based on whethe the

objeccarriesjts own energy supply orfeceiyes continuous external

power Souce. Battery-powered things carrbe moved more power tromline-power

a ed
objects. However, batteries limit the lifetime and easily than the objectss
amount of energy that the
allowed to consume, thus
driving transmission range and
Mobile 9atic: his cassification is based on frequency.
always/stay, the same locationj A sensor whether the fthing" should(nove)
should (
one objecr to anothe (for example.a may bamobila because tis mo hi
vjscosity
plant) becausd it is attached)to a
a chemical pla. or sensor moved fromom batch to
i
object (for example
loca-
tion sensor on moving
sor on
goods ina warehouse onmoving
factory the
movement may also vary, from occasional to floor)\The frequenfro
a few inches to
miles away) often
drives the permanent. The rangenge of mobility
possible power source.
 A Simplified loT Arch1tecture 41

It is not the intention of this book to endorse any one specific loT architectural
framework. In fact, it can be noted thatpromote
or

on the
architectures may differ somewhat
loT
industry use case or technology being deployed. and each has merit in
depending
the loT
heterogeneity problem discussed earlier. Thus, in this hook we present solving
framework that highlights the fundamental an loT
huilding blocks that are crmmon to most loT
systems and which is intended to
help you in designing an loT network. This framework
ispresented as two parallel stacks: The loT Data Management and Compute Stack
Core loT Functional Stack. Reducing the framework and the
in no
down to a
pair of three layer stack
way suggests that the model lacks the detail
loT necessary to develop a
strategy. Rather. the intention is to sophrsticated
sinnplify loT architecture into its most basic
building blocks and then to use it as a foundation
the
to understand key
ment principles that are applied to design and deploy
plex models are still covered, but industry-specific use cases. All the layers of mere
to they are grouped here in funetional blocks that are com
understand)
Figure 2-6 illustrates the simplified lol model easy
presented
in this honk

Core loT
Functional Stack loT Data Management
and Compute Stack

Applications
Cloud

Communications
Network Fog

Things: Sensors and

Actuators Edge

Figure 2-6 Simplified loT Architecture

Nearly every published IoT model includes core layers similar to those
left side of Figure 2-6, shown on the
including "things," a communications network, and
However, unlike other models, the framework presented applications.
data management into parallel and here separates the core loT and
aligned stacks, allowing you to carefully examine the
functions of both the network and the applications at each stage of a complex loT
This separation gives you better system.
visibility into the functions of each layer.
The presentation of theCore loT Functional Stack JIn three layers is meant to
youf understanding of thé loT architecture nto its most foundational buildingsimplity
blocks. )
imalO f course, such asimple architectureneeds to beexpanded on. The network communica-
Ao tions layerof theloT stackitselinvolvesa significant amount of detail and incorporates
a vasarray of technologiessonsider for a moment the heterogeneity of loT sensors
and the many difterent ways that exist to connect them to a network. The network
communications layer needs to consolidate these tagether, offer gateway and backhaul
technologies, and ultimately bring the data back to a central location for analysis
and processing.
 and Design
42 Chapter 2: l0T Network Architecture

Many of the last-mile technologies used in loT are chosen to meet the s
ments of the endpoints and are unlikely to ever be seen in the IT domain. H Specific req
q
network between the gateway and the data center is composed mostly af

technologies that experienced IT professionals would quickly recognize. The

tunnelingand VPN technologies, IP-based quality of service (QoS), convene
oS), ncl
routing protocols such as BGP and IP-PIM, and security capabilities such convent ional Lat
uch as encryptika
access control lists (ACLs), and firewalls.
Unlike with most networks, the applications and analytics layer of loT
IT
necessarily exist only in the data center or in the cloud. Due to the doesn't
and requirements of loT, it is often necessary to unique ch
deploy applicatic
throughout the architecture in a tiered approach, allowing data
cations and data mar
enges
and collection. anal.
intelligent controls at multiple-.points in the IloT system. In the
this book
data managementiaalignedwith each)of the three layersmodel nree

Functional Staek- The three data aanagement layers are the edge of the Core1lt
within the sensors themselves), the fog layer (data layer (data man
network), and the.cloud layer (data management inmanagement
the
in the
gateways and1agemenn
loT Data
Management and Compute Stack is examinedcloud or central data centerit
ter.Figure 2-7 highlights an éxpanded view of in greater detail later in this The
the lol architecture presented in a
this bo
Chapters 9-15 Vertical Specific Apps
(SCADA, MDMs,etc.)
Core loT
Chapter 7 Analytics Functional Stack loT Data Management
and Compute Stack

Chapter 6loT Network Mgmt Applications Cloud

(COAP, MQTT)
Chapter 5NeworkTransport Communications
Layer Network Fog
Chapter 2sateways and Things: Sensors
Backhaul Network and Actuators
Edge
Chapter 4 ACOOSs Network (FAN,
LoRa, PLC)
Figure 2-7 Chapter 3
Expanded View of the Simplified loT
As shown
in Figure 2-7, he Core JoT Architecture
containinggreater
tions layer is
Functional Stack can
detail and(specifichetwork beexpanded into
sublay
broken down into four
backhaul, IP transport, and functions. For example,
eparate sublayer the access the unicas and
The applications layer of operations and management sublayers
network. Ba
a typicalenterprise loT networks is quite
different from the
involves a _trong bignetwork.Instead
data analytics ofSimply applicatioften
using business applicat
component. One message age that is stressed
tna
throughout
 OmputO Stack 63
air

cycl conditioning
ycles can be pulses can be
smooth the coordinate
coordinated with
tated with your neighbor's,
energy demand the app ppliances in your house and your washing machine
in the
Efficiency applies to also spikes on the grid. neighborhood to
communicate to regulate theM2M commununicatións. In mining environments,
trucks, for flows betwec drills, vehicles
neeas ln
example, making sure
it. draglines,
that dump truck is
veen
bulldozers, and dump
can

smart cities, a «
always available when bulldozer
automatically by public vehicles communicate, A traffic jam is detccted and a

number oftransportation,
or

temporarily rerouteanticipatea
egulate the and the system can
quantity, instantaneous or buses servicing a specific line based buses
on traffic and
Part l of this
learned over trending. customer
book provides
ne lessons learned are detailed examples of how loT is
always shaping specific
efficiency over time. New that architecting industries.
open loT systems allows for increased
the upcoming years. When applications
and possibilities for an loT
system will appear in
for building
system open the possibility of newan loT
network, you should make sure to keep the
smart objects and more
traffic on the system.
loT Data Management and
con One or
JOne of the key
Compute Stack
messages in the first two chapters of this book is that the masSIve
of loT networks is
fundamentally driving new
architectures. For instance, Figure 1-2 in
Chapter 1 illustrates how the "things" connected to the Internet are
exponentially, with a prediction by Cisco that by 2020 there will be continuing
to
more than 50 grow
bil-
lion devices connected to some
form of an IP network. Clearly, traditional IT networks
are not
prepared for this magnitude of network devices. However, beyond the network
architecture itself, consider the data that is generated by these devices. If the number of
devices is beyond conventional numbers, surely the data generated by these devices must
also be of serious concern.
In fact, the data generated by loT sensors is one of the single biggest challenges in build-

ing an loT system. In the case of modern IT networks, by the data sourced a computer
the client/server communications model, and it serves
or server is typically generated by
sensor networks, vast majorityof data generated is the
the needs of the application. In
use on its own. For example, the majority of data generated
unstructured and of very little
more than polling data;
the communications system simply
smart meter is nothing its own
by a is still active. This data on
connection to the meter
determines whether a network data read by the meter
value of a smart meter is the metering
value. The real raw polling data
froma different
is of very little However, if you look at the
(MMS). a utility may
have millions
management system useful. For example,
be very show
the
intormation can
of the smart grid start to
perspective, If whole sections with
area. combined
entire service be analyzed and
covering its this data can
provide
of meters
to the
meters,
in the grid, to
of connectivity electrical demand
and whether the
an interruption weather reports determine
of data, such
as intormation can help
other
sources

what is happening.
This
whether s o m e otherproblem
of power o r quickly
a complete picture meters is truly
a loss
of this data can help theutility fashion.
to the analytics
of
timely
disruption in
connection a
loss of Moreover,
repair the
developed
in the grid. service outage
and

has of the extent

d e t
the
e r m i n e
 Architecture
and Design
64 Chapter 2: lOT twork

smart
object.
A natural l
ocation for
outside the to the cloud and data
location is connect

In most cases, the processing need to

cloud. Smart
objects Objectsjustneed
this processing activityi s the model is simplictty. over all
all the lot
of this has visibility
over loT
One advantage
processing is entralized. That application future.
application. in the
c t to a central cloud
needed todayand
nodes and can process all the analytics objectsconner
objects connect.
variety of
As data volume,
the ents appea,
limitations. requiremen
However this model also has new increase,

closertothe oT system.
loT
ing to the network, and theneed fo more efficiency
dataanalysis
and those requirements (tendto bring he need
lor
These new requirements include the following:
such
of industrial systems)
matter for many types
shutdowns orrestoreelectrical
latency:(Milliseconds res
inimizing line
as when you are trying to prevent manufacturing the data can
make a differ.
that collected
Service. Analyzingdata close to thé device
failure.
ence between/averting disaster and a
cascading system
Offshoreoilrigsgenerate 500 GB ot dataweekly
Onserxing networkbandwidth: It isnot practical/to
10 TB for every 30minutes flight.
of
ommercial generate
jets or hundreds ofthousands ofedge
ransport vastamounts of data from thousands
critical analysesdo not
devices to thecoud Noris itnecessary because many
require Cloud-scaleprocessing and storage.
a wide geographic
Increasing local eficiency<Collecting and securing data across
areawith different environmental condition_ may not be useful. The environmental
conditionsinone area will (rigger a local responseindependent from the conditions
of anothetsite Jundreds of miles away.Analyzing both areas in the same cloud
sys-
tem may not be necessary for immediate efficiency.

An important design consideration, therefore, is how to design an loT network to manage

this volume of data in an efficient way such that the data can be
quickly analyzed and
lead to business benefits. The volume of data generáted by loT devices can
be so great
that it can easily overrun the capabilities of the headend system in the
data centeror the
cloud. For example, it has been observed that a moderately sized
of 1 million meters will generate close to 1
smart meter rietwork
reads and other instrumentation
billon data points each day (including
meter
dataj, resulting inofl 18,ot data. For an IT organization
that is not prepared to contend with this volume data storage
this creates a whole new challenge.
and real e analysis,
al-time
The volume of data also introduces questions about bandwidth
sive amount of loT data begins to funnel into the data
of traffic? Does the
center, does themanagement.As the mas-
this volume
capacity to sustain
application network
have the
ingest, store, and analyze
the vast quantity of data
that is server have
the ability
referred to as the "impedance
mismatch" of the
data coming in? This is to
ability to deal with that data generated by the loT sometimes
management application's data. system and the
As illustrated
anagement traditional
in Figure 2-14, data management in trad..
IT
(laptops, printers, phones, and so on)
IP
The endpoints
loud. systems is very
work to servers
in the data
center or cloud. Data i
communicate
Data is
generally
re typicall. 5enerally
over an IP simple.
links from
a c c e s s to core
are
typically high stored in the core
net
the physical
data is quick.
bandwidth, data
center,
meaning access to and
IT
 lOT Data Management and Compute Stack 65

Data Center/Cloud

Core Network

Tellable-

Endpoints

Figure 2-14 The Traditional IT Cloud

Computing Model
loT systemsfunction differently
Several Gata-related problems need to be addressed:
Bandwidthin last-mile IOT networks is verKlimitea)
millions of
devices.available bandwidth when dealing with thousands/
or even less. may be onorder of tensof Kbps per device
Latency can be very high. Instead of
ldealing with latencyin the milliseconds
range,large loT networks(often introducy
milliseconds. latency of (hundreds to thousandy of
Networkbackhaul jrom the gateway can be unreliable and often
oreven satellite links.
Backhaul links can also be expensive if depends on 3G/LTE
model is necessary. 4per-byte data usage
The volume of
data transmitte pver the backhaul can be high, and much of the data
may not really be that
interesting (such as simple polling messages).
Big data is getting bigger. Theconcept ofstoring
and analyzingall stasor datajn the
cloud is tmpractical) Thd sheer vounme bf data generated
fesponserO the data almostumpossible. makes(çal-time analysis and

FogComputing
The solution to the
challenges mentioned in the previous section is to distribute data
management throughout the loT system, as close to the edge of the
i? petwork as
sible.The best-known embodiment ofedge services in loTtog computing. Any pes
with computing, storage, and network Eonnectivit) can be( fog node device
Pxamples îmtcturde
industrial controllers, switches, foutèrs) embedded servers) and tor
loT data close to where it is collected minimizes
gateways Analyzing
latency, offloads gigabytes of network
traffic from the core network, and keeps sensitive data inside the local network.
 1

O0Chapter 2 loT Netwok Architocture an1 Dgj

Bonomi and Rodolfo Milito oe

of fog was first developed
by Flavio relative to com
OTe The concept name from a
comparison
CIsco Systems, In the world of loT, fog
gets its near the ground, In the
rests
clouds exist in the sky, fog
puting in the cloud layer. Just as
r e s o u r c e s as
close to the ground
is to place
way. the intention of foK Computing is that
the term "fog" was
ne An interesting
side note
n a t s, the loT devices-as possible. not working directly in lot
Rodolfo's wife.
Although
actually coined by Ginny
Nichols, and was able to quickly
husband was developing
of what her made the sugge
she had an excellent grasp One day she
between cloud and edge computing.
raw the comparison The name stuck.
tion of simply calling it the "fog layer."

gathering (such as
is that the fog noddallows ntelligence it allows
An advantage of this structure and in dong so, better
analytics) andcontrol Jrom
the closest possible point, introduces a new layer to the
dne sense,
this
constrained-networks, In
pertormanee-ever referred to as the "tog layet] Figure 2-13
one that is often
traditionakIT computing model and Compute Stack
in the loT Data Management
shows the placèment.pf the fog layer

Hundreds Data Center/Cloud

Datà Center/Cloud
IRT
Transactional Response Times

Core IPv6 Network

Thousands
Backhaul)

oloo 010101
010101
11001
Ca
001
11001
Tens of Thousands 001 Fog Layer
sensing
Multi-Service Edge
Millsecond/Sub-Second Response

Millions
Embedded Systems uo Smart Objects
and Sensors
Low Power
LowBandwidth
and Compure Stack wirh Fog Compuring8
Figure 2-15 The loT,Data Management
to

éccomplished)veryclose to
the édge devic sittingas closehas
Fogservices aretypically this is that thetog nod
possiblejQne significant advantage of
theloT endpointsas because of its geographic proxim
the Sensors it is managing
contextual awarenesSot an oil derricktha
example, there might be a fog router.on
ity to those sensors.
For
at that location. Because
the fog node is able to anay
sensor activity
monitoring all the
 Or Stack 67
loT Date Management and Compute

Cinformationfrom all the sensors on that derrick, itlkan providoptextual analysi5p

the messages it is receiving and maydccideto send backonytherelevant information

over the backhaul networh to the cloud, Ih this way, it i_pérforming distributed anaiytics
such that the volume of datá sentupstream is greatyreduced and is much ore use
application and analytics servers
residing in the ¬loud.
In addition hasing contextual awareness gives foghodesthe ability toteact to events)
in
the loT networ much moreguickly than in the(traditional IT compute modewnicn
Would likely incur greater
latency and have slower response times. The fog layer thus con
provides a distributed edge control loop capability, where devices can be monitored,
trolled, and analyzed in real time without the need to wait for communication rrom
central analytics and application servers in the cloud.
The value of, this model is clear. For example, tire pressure sensors on a large truck in an
open-pit mine might continually report measurements all day long. There may be only
minor pressure changes that are well within tolerance limits, making continual report
ing to the cloud unnecessary. Is it really useful to continually send such data back to the
{ cloud over a potentially expensive backhaul connection? With a fog node on the truck,
it is possible to not only measure the pressure of all tires at once but also combine this
data with information coming from other sensors in the engine, hydraulics, and so on.
With this approach, the fog node sends alert data upstream only if an actual problemnis
beginning to occur on the truck that affects operarional efficiency/
loTfog computing enables datato bepreprocessed and.correlatedwith other inputs to
producereleyant informationjThisatá can then beused ésreal-time,actionable knowl-
o edgebf loT-enabled)applicationsJLonger term, this data cànbeused to gain a deeper
understanding of network behavior and systems for the purpose of developing proactive
policies, processes, and responses.
Fog applications are as diverse as the Internet of Things itself. What they have i
mon is data reduction-monitoring or analyzing real-time data from network-connected
equipment settings,
things and then initiating an action, such as locking a door, changingvalve in response to a
applying the brakes on a train, zooming a video camera, opening
a

pressure reading, creating a

bar chart, or sending an alert to a technician to make a pre
ventive repair.
follows
LThe definingharacteristic of fog computing àre
as

The fog node sits as close to the

Contextual locationawareness andlow latency:
deliver distributed computing
loT endpoint as possible to
more centralized cloud, the
In sharp contrast to the
Geographic distribution: demand widely distributed
services and applications targeted by the fog nodes
deployments.
in the presence
loT endpoints: Fog nodes are typically deployed
Deployment near For example, typical metering
deployments
number of loT endpoints. as the fog
of large
a 4000 nodes per gateway router,
which also functions

often see 3000 to

computing node.
 Cornpute Stack 69
loT Data Management ang
Gra
analyzing and otherwise anaging úatayndpoints. This faves the clouct frem being queted
by each and everynoo for each¬vend cd..>
LThis model suggests adhierarchical organization of net work, (omputand dataatorag)
resourceskAt each stage, data is colected, analyzed, and responledto when necessary.
according to the capabilities of
the resourcesatenehlayer. As data needs to be sent to
the cloud, the latency becomes higher. The (dvantagh this hicrarchy is that aCEsponse
oeventS from fesourçesjclose to the end devic Is last nd can result in immediate ben-
efits, while still having deeper compute respurces available in the cloud when necessary.

Ut is important to note that theheterogeneity of

loTdevices lso meang a
of edge and tog computing résources. While cloud resources are expectedheterogene
to be homoge
nous, it is tair to expect that may cases both edge and fog resources will sse different
operating systems, have different CPU and data storage capabilities, and have different
enerey consumption profilesEdge and fog thus fequirean abstraction laye that
applications to communicate with one anothéTheabstraction layer eposesbcommon
allows

setof APls for monitoring, provisioning, and controlling the physicaresourcesn a stan-
dardized wayThe abstraction layer alsoequires a mechaniswto supportvirtualization,
with
the ability tarun multiple operating systemspr serviceeentainers-an-physical
devices to suppork multitenancy and application(conaístenc) açross the loT system.
Definition of a common communications servicestramework is being addressed by
E goups such as oneM2M, discussed earlierlFigure 2-16 illustratesthe hierarchical natur
6 4 of edge, fog, and cloud computing across an IoT system.J

Applicatlon "B_Cloud"
High Latdncy Cloud Layer
Applicatlon "A_Cloud
(Core Data Center)
Os OS OS Os
CVMC VM CVMCVM
Databas
OS
OS
VM,
VM
Applilcationa Cor
LOS OS
CVM C VM
Aanliaatlan"A Faa"
Eog Computing Leyer Application "A_Fog" AppllcationO Pog
CMultServic Edge
(loT Gateway 9 6o s COs O s
VM M CVM O VM
End-Polnt Appllcetion B
End-Polnt Appllcation A

Edpe Layer
Embedded Systems and OS
OS OS
SensorsVehicles
Machines, etc) CVMC VM

Across a n
loT System
Low Latency Management
and Data
Distributed
Compute
Figure 2 - 1 6