04-Sep-21

Chapter 2

IoT Architecture

1
 04-Sep-21

IoT is consists of the following components

IoT Components
 Things we connect : Hardware, sensors, actuators
 Connectivity : Medium we use to connect things
 Platfrom : processing and storing collected data
 Receive and send data via standardized interfaces or
API
 Store the data
 Process data
 Analytics
 Get insights from gathered data
 User interface

2
 04-Sep-21

IoT Devices
 The "Things" in IoT usually refers to IoT devices which
have unique identities
 They can perform remote sensing, actuating and monitoring
capabilities
 IoT devices can:
 Exchange data with other connected devices and
applications
 Collect data from other devices and process the data
locally
 Send the data to centralized servers or cloudbased
application back-ends for processing
 Perform some tasks locally and other tasks within
the IoT infrastructure

Generic Block Diagram of IoT Device

3
 04-Sep-21

SMART OBJECTS: THE “THINGS” IN IoT

 Sensors
 Characteristic of any device or material to detect
presence of particular physical quantity
 Output of sensor is signal, which is converted to
human readable form
 Performs some function of input by sensing or feeling
physical changes in the characteristic of a system in
response to stimuli
 Input: Physical parameter or stimuli
 Example: Temperature, light, gas, pressure, and sound
 Output: Response to stimuli

Sensors

4
 04-Sep-21

Classification

 Passive Sensor
 Cannot independently sense the input
 Example:
 Accelerometer
 Soil Moisture
 Water-level and
 Temperature Sensors
 Active Sensor
 Independently sense the input
 Example:
 Radar
 Sounder and
 Laser Altimeter Sensors

 Analog Sensor
 The response or output of the sensor is some continuous
function of its input parameter
 Example:
 Temperature sensor

 LDR

 Analog Pressure Sensor and

 Analog Hall Effect/Magnetic Sensor

 Digital Sensor
 Responses in binary nature
 Designs to overcome the disadvantages of analog sensors
 Along with the analog sensor it also comprises of extra
electronics for bit conversion
 Example:
 Passive Infrared (PIR) Sensor and

 Digital Temperature Sensor (DS1620)

5
 04-Sep-21

 Scalar Sensor
 Detects the input parameter only based on its
magnitude
 Response of the sensor is a function of magnitude of
the input parameter
 Not affected by direction of input parameter
 Example: Temperature, Gas, Strain, Color and Smoke
Sensors
 Vector Sensor
 Response of sensor depends on magnitude of direction
and orientation of input parameter
 Example :
 Accelerometer, Gyroscope, Magnetic Field
 Motion Detector Sensors

Actuator

 Part of system that deals with the control action

required (mechanical action)
 Mechanical or electro-mechanical devices

6
 04-Sep-21

Actuator

 A control signal is input to an actuator and an

energy source is necessary for its operation
 Available in both micro and macro scales
 Example:
 Electric Motor,
 Solenoid,
 Hard Drive Stepper Motor
 Comb Drive
 Hydraulic Cylinder,
 Piezoelectric Actuator
 Pneumatic Actuator

Actuator - Classification

7
 04-Sep-21

Smart Objects

Bell’s Law of Computer Classes:

A new computer class emerges roughly every decade

“Roughly every decade a new, lower

10T
Mainframe priced computer class forms based on
1 per Enterprise a new programming platform,
) computer)

1T
network, and interface resulting in
100G new usage and the establishment of a
10G new industry.”
[Bell et al. Computer,
1G 1972, ACM, 2008]
100M Workstation
Size (mmper

10M 1 per Engineer

3

1M Smart
log (people

100k Laptop Sensors

10k 1 per
Professional
1k
Mini
100 Computer
10 1 per Company
Personal 1 per Family 1 per person 100 – 1000’s
1
Computer Smartphone per person
100m
1950 1960 1970 1980 1990 2000 2010 2020

16

8
 04-Sep-21

IoT analytics

17

Cisco

18

9
 04-Sep-21

BUILDING BLOCKS of IoT

 Four things form basic

building blocks of the IoT
system –sensors, processors,
gateways, applications. Each
of these nodes has to have its
own characteristics in order
to form an useful IoT
system.

Sensors:
 These form the front end of the IoT devices. These are the
so-called “Things” of the system. Their main purpose is to
collect data from its surroundings (sensors) or give out data
to its surrounding (actuators).
 These have to be uniquely identifiable devices with a
unique IP address so that they can be easily identifiable
over a large network.
 These have to be active in nature which means that they
should be able to collect real-time data. These can either
work on their own (autonomous in nature) or can be made
to work by the user depending on their needs (user-
controlled).
 Examples of sensors are gas sensor, water quality sensor,
moisture sensor, etc

10
 04-Sep-21

Processors:

 Processors are the brain of the IoT system. Their

main function is to process the data captured by the
sensors and process them so as to extract the
valuable data from the enormous amount of raw
data collected. In a word, we can say that it gives
intelligence to the data.
 Processors mostly work on real-time basis and can
be easily controlled by applications. These are also
responsible for securing the data – that is
performing encryption and decryption of data.
 Embedded hardware devices, microcontroller, etc
are the ones that process the data because they have
processors attached to it.

Gateways:

 Gateways are responsible for routing the

processed data and send it to proper locations for
its (data) proper utilization.
 In other words, we can say that gateway helps in
to and fro communication of the data. It provides
network connectivity to the data. Network
connectivity is essential for any IoT system to
communicate.
 LAN, WAN, PAN, etc are examples of network
gateways.

11
 04-Sep-21

Applications:

 Applications form another end of an IoT system.

Applications are essential for proper utilization of
all the data collected.
 These cloud-based applications which are
responsible for rendering the effective meaning to
the data collected. Applications are controlled by
users and are a delivery point of particular
services.
 Examples of applications are home automation
apps, security systems, industrial control hub,
etc.

Elements of IoT

12
 04-Sep-21

The elements and key technologies of IoT.

IoT Elements Technologies

Naming Electronic, Product Code, Ucode
Identification
Addressing IPv4, and IPv6
Smart, Sensors, RFID Tags, Wearable Sensing
Sensing
Devices and Actuators
Radio Frequency Identification, Wireless
Sensor Network, Near Field
Communication
Communication (NFC), Bluetooth, Long Term
Evolution (LTE)
Hardware Audrino, Raspherry Pi, Intel Galil
Computation
Software Operating System
Identity-Related, Information Aggregation,
Services
Collaborative-Aware and Ubiquitous
Semantics RDF, OWL, EXI

The IoT Architectural landscape

 Thousands of new applications spanning

numerous domains.
 Each comes with its own requirements;
combining these leads to complex (difficult
to manage) and often proprietary systems.
 Defining a unified architecture is challenging
and interoperability problematic if too many
standards to chose from
 Documentation scattered and often difficult
to navigate.

13
 04-Sep-21

IoT Architecture

 An IoT architecture consists of your

networked things, typically wireless sensors
and actuators.
 It includes sensor data aggregation systems
and analog-to-digital data conversion
 It includes the processing and business .The
five layers are perception, transport, processing,
application, and business layers.

14
 04-Sep-21

How do IoT Work?

 An IoT system consists of sensors/devices

which “talk” to the cloud through some kind
of connectivity. Once the data gets to the
cloud, software processes it and then might
decide to perform an action, such as sending
an alert or automatically adjusting the
sensors/devices without the need for the user

15
 04-Sep-21

The three-layered architecture of IoT.

 It is a very basic architecture and fulfills the

basic idea of IoT

Perception Layer (Lớp cảm nhận)

 It is also known as a sensor layer.

 It works like people’s eyes, ears and nose.
 It has the responsibility to identify things and
collect the information from them.
 There are many types of sensors attached to objects
to collect information such as RFID, 2-D barcode
and sensors.
 The sensors are chosen according to the
requirement of applications.
 The information that is collected by these sensors
can be about location, changes in the air,
environment, motion, vibration, etc.

16
 04-Sep-21

Network Layer

 Network layer is also known as transmission layer.

 It acts like a bridge between perception layer and
application layer.
 It carries and transmits the information collected
from the physical objects through sensors.
 The medium for the transmission can be wireless or
wire based.
 It also takes the responsibility for connecting the
smart things, network devices and networks to each
other.

Application Layer

 Application layer defines all applications that use

the IoT technology or in which IoT has deployed.
 The applications of IoT can be smart homes,
smart cities, smart health, animal tracking, etc.
 It has the responsibility to provide the services to
the applications. The services may be varying for
each application because services depend on the
information that is collected by sensors.

17
 04-Sep-21

Four Layer Architecture

 The three-layer architecture was most basic architecture.
Due to continuous development in IoT, it could not
fulfill all the requirements of IoT. Therefore,
researchers proposed an architecture with four layers

Support Layer

 The reason to make a fourth layer is the security in

architecture of IoT
 Information is sent directly to the network layer in
three-layer architecture.
 Due to sending information directly to the network
layer, the chances of getting threats increase. Due to
flaws that were available in three-layer architecture,
a new layer is proposed.
 In four-layer architecture, information is sent to a
support layer that is obtained from a perception
layer. The support layer has two responsibilities.
 It confirms that information is sent by the authentic users
and protected from threats

18
 04-Sep-21

Five Layer Architecture

 The four-layer architecture played an important role

in the development of IoT.
 There were also some issues regarding security and
storage in four-layer architecture. Researchers
proposed five-layer architecture to make the IoT
secure
 It has three layers like previous architectures whose
names are perception layer, transport layer and
application layer.
 It also has two more layers. The names of these
newly proposed layers are processing layer and
business layer

Processing Layer

 The processing layer is also known as a middleware

layer
 It collects the information that is sent from a
transport layer. It performs processing onto the
collected information.
 It has the responsibility to eliminate extra
information that has no meaning and extracts the
useful information
 However, it also removes the problem of big data in
IoT. In big data, a large amount of information is
received which can affect performance of IoT. T

19
 04-Sep-21

Business Layer

 The business layer refers to an intended behavior

of an application and acts like a manager of a
whole system.
 It has responsibilities to manage and control
applications, business and profits models of IoT.
The user’s privacy is also managed by this layer.
 It also has the ability to determine how
information can be created, stored and changed.

More details layer

20
 04-Sep-21

Application Layer

 The application layer defines all applications in

which IoT has deployed
 IoT Applications such as:
 Smart home
 Smart health
 Smart cities
 It has the authority to provide services to the
applications.
 The services may be different for each
application because of services based on the
information collected by sensors.

Data Processing Layer

 In three-layer architecture, the data were directly

sent to the network layer
 In four-layer architecture, data is sent to this
layer that is obtained from a perception layer.
 Data Processing Layer has two responsibilities it
confirms that data is forwarded by the authentic
users and prevented from threats.

21
 04-Sep-21

Network Layer

 This layer is also known as a transmission layer.

 It acts like a bridge that carries and transmits
data gathered from physical objects through
sensors.
 The medium can be wireless or wire-based.
 It also connects the network devices and
networks to each other

Perception layer/Sensor layer

 The sensor layer has the responsibility to

recognize things and gather the data from them.
 There are many types of sensors connected to the
objects to gather information such as RFID,
sensors and 2-D barcode
 The sensors are selected as per the requirement
of applications
 The data that is collected by these sensors can be
about location, changes in the air, environment,
etc.

22
 04-Sep-21

Architecture

 An architecture (of a system) is

“The fundamental organization of a system
embodied by its components [jl:building blocks],
their relationships to each other [jl: connectors
and interfaces,dependencies] and to the
environment and the principles guiding its design
[jl:rationales for choices, rules & constraints for
building blocks and connectors]
and evolution”

 An architecture description is
 a collection of models organized into views
that examine a system from a certain
viewpoint defined by the concern of a
stakeholder
 for understanding, analysis, communication,

construction, documentation

23
 04-Sep-21

Main design principles

 Main design principles and needed capabilities

 The approach taken in SENSEI was to develop an
architecture and technology building blocks that enable
a “Real World integration in a future Internet.”
 The telecommunications industry, meanwhile, has
focused on defining a common service core for
supporting various M2M applications
 The approach taken in IoT-A differs from the two
approaches above in the sense that instead of defining a
single architecture, a reference architecture is created,
captured in what the IoT-A refers to as the
Architectural Reference Model (ARM).

 Design for reuse of deployed IoT resources across

application domains
 Design for different abstraction levels that hide
underlying complexities and heterogeneities.
 Design for sensing and actors taking on different
roles of providing and using services across different
business domains and value chains.
 Design for ensuring trust, security, and privacy.
 Design for scalability, performance, and
effectiveness.
 Design for evaluability, heterogeneity, and simplicity
of integration

24
 04-Sep-21

 Design for simplicity of management.

 Design for different service delivery models.
 Design for lifecycle support. The lifecycle phases
are: planning, development, deployment, and
execution. Management aspects include deployment
efficiency, design time tools, and run-time
management.

One M2M IoT Standardized Architecture

25
 04-Sep-21

 The one M2M architecture divides IoT functions

into three major domains:
 Application Layer
 Services Layer
 Network Layer
 Application Layer
 The one M2M architecture gives major focus on
connectivity between devices and their applications
 Includes the application-layer protocols
 Attempts to standardize northbound API definitions
for interaction with BI systems

 Applications tend to be industry-specific and have

their own sets of data models
 Thus shown as vertical entities
 Services Layer
 Shown as horizontal framework across the vertical
industry applications
 At this layer, horizontal modules include
 Physical network that IoT applications run on
 Underlying management protocols
 Hardware
 Examples include backhaul communications via
cellular, MPLS networks, VPNs, and so on

26
 04-Sep-21

 Services Layer
 Riding on top is the common services layer.
 This conceptual layer adds APIs & middleware
supporting third-party services & applications
 Network Layer
 Communication domain for the IoT devices and
endpoints
 Includes devices themselves and the communications
network that links them
 Examples include wireless mesh technologies, such as
IEEE 802.15.4, and wireless point-tomultipoint
systems, such as IEEE 801.11ah

IoT World Forum (IoTWF) Standardized Architecture

27
 04-Sep-21

 Defines set of levels with control flowing from

the center (cloud service or dedicated data
center), to the edge
 May includes sensors, devices, machines, and
other types of intelligent end nodes
 In general, data travels up stack, originating
from edge, and goes northbound to the center
 Using this model, we are able to achieve:
 Decompose the IoT problem into smaller parts
 Identify different technologies at each layer and
how they relate to one another

 Define a system in which different parts can be

provided by different vendors
 Have a process of defining interfaces that leads to
interoperability
 Define a tiered security model that is enforced at
the transition points between levels

28
 04-Sep-21

Layer 1: Physical Devices and Controllers Layer

 This layer is home to the “things” in the IoT,

including various endpoint devices & sensors
 Size of these “things” can range from almost
tiny sensors to huge machines in factory
 Their primary function is generating data and
being capable of being controlled over network
 Sensor gathers the records however that we need
to convert it into understandable format & join the
one’s sensor device using some protocol that we
want to configure right here in layer two

Layer 2: Connectivity Layer

 Communications Between layer 1 devices
 Reliable delivery of information across the network
 Switching and routing
 Translation between protocols
 Network level security
 Network connectivity, join your tool with wi-fi
connectivity or net stressed connection. This
connectivity is changed based on context & area.

29
 04-Sep-21

Layer 3: Edge Computing Layer

Upper Layers: Layers 4–7

30
 04-Sep-21

A Simplified IoT Architecture

A Simplified IoT Architecture

 Framework is presented as 2 parallel stacks:

 IoT Data Management and Compute Stack
 Core IoT Functional Stack
 Intention is to simplify the IoT architecture
into its most basic building blocks

31
 04-Sep-21

A Simplified IoT Architecture

The Core IoT Functional Stack

 Every published IoT model include core layers, including

“things,” a communications network, and applications
 Framework presented here separates core IoT & data
management in parallel & aligned stack
 Allows you to carefully examine functions of both,
network & applications, at each stage of complex IoT
system
 This separation gives you better visibility into the
functions of each layer
 Presentation in 3 layers is meant to simplify your
understanding of IoT architecture into its most
foundational building blocks
 Such simple architecture needs to expanded

32
 04-Sep-21

IoT Data Management and Compute Stack

 The “things” connected to the Internet are

continuing to grow exponentially
 Cisco predicted that by 2020 there will be
more than 50 billion devices connected to some
form of an IP network
 If number of devices is beyond conventional
numbers, surely the data generated by these
devices must also be of serious concern


IoT Data Management and Compute Stack

33
 04-Sep-21

IoT Data Management and Compute Stack

 Data-related problems need to be addressed:

 Bandwidth in last-mile IoT networks is very limited
 Latency can be very high (Instead of dealing with
latency in the milliseconds range, large IoT networks
latency of hundreds to thousands of milliseconds)
 Volume of data transmitted can be high
 Big data is getting bigger

Fog Computing
 Solution to the various challenges is to distribute data
management throughout the IoT system, as close to the edge of
the IP network as possible
 Best-known example of edge services in IoT is fog computing
 Any device with computing, storage, and network connectivity
can be a fog node
 Concept of fog was first developed by Flavio Bonomi and
Rodolfo Milito of Cisco Systems
 In world of IoT, fog gets name from a relative comparison to
computing in cloud layer
 Like clouds exist in sky, fog rests near ground
 In the same way, the intention of fog computing is to place
resources as close to the ground—that is, the IoT devices—as
possible

34
 04-Sep-21

Fog Computing

 Examples : industrial controllers, switches,

routers, embedded servers, and IoT gateways
 An advantage of this structure is that fog node
allows intelligence gathering (analytics) and
control from the closest possible point
 In one sense, this introduces new layer to the
traditional IT computing model, one that is often
referred to as the “fog layer”


Fog Computing

35
 04-Sep-21

Edge Computing

 Also called as “mist” computing

 If clouds exist in sky, and fog sits near ground,
then mist is what actually sits on the ground
 Thus, concept of mist is to extend fog right
into IoT endpoint device itself
 Fog computing solutions are being adopted by
many industries

Cloud vs Fog vs Edge

 Cloud computing dominated the networked

systems landscape until recently
 End-devices merely information gatherers
 All intelligence in the cloud (relational DBs, analytics,
web interfaces, control functions)
 As the number of devices grow, applications
diversify and generate more data, this will not
scale
 Routing and storage costs
 Signalling overheads
 Latency inappropriate for real-time apps

36
 04-Sep-21

Ultimately it is about performance

Fog computing

 Pushing some of the effort closer to the device,

i.e. to access networks/gateways.
 This includes data aggregation, compression,
(partial) processing; making localised decisions.
 IoT device
 Not required to be extremely smart, i.e. unique
address and ability to communication directly with
the cloud
 No substantial storage
 Battery powered (lifetime)

37
 04-Sep-21

Example: Home automation

Example: Fitness tracking systems

 The user’s mobile phone acting as gateway

38
 04-Sep-21

Edge computing
 (process as much as possible where data is collected)
 Pushing processing power, communication
capabilities, intelligence down at device level
 Emerging applications range from autonomous
vehicles, to VR glasses, to earbuds.
 Do as much processing as required on the device, transmit
only what is relevant long term or summaries
 Low latency and decentralised decisions

 Less signalling and communication overhead

 Personalised experience

Example: Deep Learning at the Edge

 Hardware Support: Low-power chips specialised

in computationally intensive tasks (IBM
TrueNorth,Movidius, Huawei Kirin)
 Software: lightweight inference frameworks
optimised for constrained devices (mobile
TensorFlow, Apple CoreML, DeepSense)
 Dedicated NN architectures: Model compression
(SqueezeNet, MobileNet), point-wise group
convolution (ShuffleNet), model pruning
(NestDNN).

39
 04-Sep-21

ABCD’s of IoT

Applications

40
 04-Sep-21

Big Data Analytics

 Map-Reduce
 Frequent Item-sets
 Similarity
 Clustering
 Dimension Reduction
 Streaming Data

Why is there big data?

 Why is there big data?

 Number of devices increasing exponenEally
 They conEnuously generate data
 For example, on average, 72 hours of videos
are uploaded to YouTube in every minute.

41
 04-Sep-21

IoT Meets Big Data

42
 04-Sep-21

Big Data in WSN/IoT

 Big Data
 Volume: size of data such as terabytes (TB), petabytes (PB),
zettabytes (ZB),
 Variety: types of data from difference sources (sensors,
devices, social networks, the web, mobile phones)
 Velocity: how frequently the data is generated (every
millisecond, second, minute, hour, day,week, month, year.)
Processing frequency may also differ from the user
requirements.
 Challenges
 High volume processing using low power processing
architectures.
 Discovery of real-time data-adaptive Machine learning
techniques.
 Design scalable data storages that provide efficient data
mining

43
 04-Sep-21

How much data is big?

 2010: Apache Hadoop:

“datasets which could
not be captured, mana
ged, and processed
by general computers
within an acceptable
scope.”
 3V model: Volume, Ve
locity, Variety [META]
-+1V: Value [IDC]

Value of Big Data

 New business and e

fficiency opportuniEs

 $300B in US medica
l industry
 Increased efficiency
of government
operaEons
 Search engines perso
nalized for users
 Personalized ads, pro
ducts, etc.

44
 04-Sep-21

IoT and Big Data

 IoT applicaEons conEnuously generate data

 Even the smallest device generates data
 The problem: data processing capacity is lower than
data generation speed

Data Generation
 Enterprise data: big companies, e.g. Facebook, Amazon
 Business data is expected to double every 1.2 years
 Walmart processes 1M customer trades/hour
 Akamai adverEses 75M events/day
 IoT data: pervasive applicaEons, clinical medical care-‐R&D
 Large scale, heterogeneous and strongly correlated data
 30 billion RFID tags and 4.6 billion camera phones are used around the
world today
 If Wal-‐Mart operates RFID on item level, it is expected to generate 7
terabytes (TB) of data every day
 Bio-‐medical data: human gene sequencing
 One sequencing of human gene may generate 100 sequences of 600GB
raw data
 Other areas: physics, bio-‐informaEcs, etc.
 Astronomy: Sloan Digital Sky Survey (SDSS), the data volume generatd
per night surpasses 20TB 8

45
 04-Sep-21

Data Acquisition

 Log files: almost all digital dev

ices provide logging capability
 Web activity recording, financia
l applications, network monitori
ng
 Sensing: physical quantities into
readable digital signals
 Sound wave, voice, vibration, a
utomobile, chemical, current, we
ather, pressure, temperature, etc.
 Localization
 Mobile plakorms: similar to sen
sing
 More personalized, specific to a
user

Data Transportation

 Data transfer to a storage infrastr

ucture for
processing and analysis
 Inter data center network (DCN)
transmissions:
 Source to data center
 Using WAN: 40-‐100Gbps
 Intra DCN transmissions:
 Data center interconnect
 Top-‐of-‐the-
‐rack vs.aggregator switches
 1-‐10-‐100 Gbps

46
 04-Sep-21

Data Preprocessing
 Eliminate or reduce redundancy, noise, meaningless data
 Increase storage efficiency, data analysis speed
 IntegraEon: combining data from different sources
 Data warehouse: ETL (Extract, Transform and Load)
 Data federaEon
 Mostly used by search engines
 Cleaning: how can data be cleaned?
 Define error types -‐> idenEfy errors -‐> correct errors -‐>
document errors -‐> modify infrastructure to prevent errors
 Redundancy eliminaEon
 Redundancy detecEon, data filtering, data compression
 Areas: Images, videos
 One solution: Compression!

Preprocessing Capabilities

 Assume there is a job with 1TB total size

 100K Arduino, 1K Raspberry Pi 2, 100 servers
 Time spent in computaEon vs. networking
 Ardunio level
 Raspberry Pi 2 level
 Server level

47
 04-Sep-21

Big Data Storage

 Storage and management of large-
‐scale data sets while achieving reliability and availability of data accessin
g
 TradiEonally on servers with structured RDBMSs.
 ExisEng storage systems for massive data
 Direct asached storage (DAS)
 Several hard disks directly connected with servers
 Only suitable to interconnect servers with a small scale
 Network asached storage (NAS)
 NAS uElizes network to provide a union interface for data access and
sharing
 The I/O burden is reduced extensively since the server accesses a storage
device indirectly through a network
 Storage area network (SAN)
 Designed for data storage with a scalable and bandwidth intensive network
 Data storage management is relaEvely independent within a storage local
area network 13

Distributed Storage System

 CAP: Consistency Availability Partition tolerance

 At most two of the three requirements can be satisfied simultaneously
 CA vs. CP vs. AP systems
 CA: for single servers
 CP: useful for moderate load [BigTable and Hbase]
 AP: useful when no high demand on accuracy [Dynamo and Cassandra]

48
 04-Sep-21

Connectivity

 M2M
 Wireless Sensor Networks
 IPv6 and 6LowPAN
 Bluetooth LE and ZigBee
 WiFi and LTE
 Backscatter

Devices and Platforms

 Mobile Systems
 Sensor Systems
 Wearables
 Energy Harvesting
 Security and Privacy

49