INTERNET OF THINGS (21CS735) MODULE 3 NOTES

Module 3: IoT Processing Topologies and Types: Data Format, Importance of Processing in IoT,
Processing Topologies, IoT Device Design and Selection Considerations, Processing Offloading.
Data format refers to how data is organized, stored, and accessed in any system, including the Internet of Things
(IoT). In the context of IoT, vast amounts of data are generated by various sources such as sensors, devices, users,
and monitoring systems. Proper data formatting helps ensure that this data can be efficiently processed, analyzed,
and utilized.
Role in IoT Applications:
 Efficiency: IoT systems generate a large volume of data. Properly formatted data enables easier
processing, efficient querying, and faster data retrieval.
 Interoperability: Different IoT devices and applications often interact with each other. Using
standard data formats ensures smooth communication between different systems.
 Storage & Management: Data formats affect how the data is stored, which influences the choice
of databases (e.g., SQL for structured data, NoSQL for unstructured data).
 Data Integrity: Well-defined formats ensure that data is captured accurately, minimizing errors and
redundancies.

Figure 6.1 - Data Sources and Types in IoT

The figure illustrates various sources of data generation and storage in IoT systems, showing how human-driven
activities and machine-generated data contribute to the massive volume of data flowing over the Internet.
Key Components:
 Internet & Cloud: These serve as the primary hubs for data storage and transfer, facilitating the
flow of information between users, sensors, and devices.

@VTUPADHAI – YT CHANNEL 1
 INTERNET OF THINGS (21CS735) MODULE 3 NOTES

 Human Data Sources: Users contribute data through emails, social media posts, videos,
transactions, etc.
 Non-human Data Sources: Sensors, buildings, traffic systems, and industries are examples of
automated systems generating real-time data, like traffic information or environmental conditions.
The diagram highlights how various data sources — both human and machine — contribute to the global IoT
ecosystem, generating different types of data that need to be processed and stored effectively.

Types of Data Format in IoT

There are two major types of data formats based on their structure:
1. Structured Data:
o Pre-defined and organized data, often stored in relational databases (RDBMS).
o Easily searchable through standard querying languages like SQL.
o Examples: Flight reservation data, banking transactions, inventory records.
o Role in IoT: Though significant in certain applications (like industrial automation),
structured data constitutes a small portion of IoT-generated data.
2. Unstructured Data:
o Data without a fixed format or structure, making it difficult to process using conventional
methods.
o Stored in NoSQL databases due to its flexibility.
o Examples: Sensor data, satellite images, surveillance footage, emails, and social media
posts.
o Role in IoT: The majority of IoT data is unstructured, such as data from sensors,
surveillance systems, or video feeds from smart devices.
Each type has its specific use case in IoT, with unstructured data being the dominant form due to the variety and
unpredictability of data generated by IoT sensors and systems.

Importance of Processing in IoT

In IoT applications, the vast amount and diverse types of data necessitate efficient and intelligent processing
techniques. With the rapid growth of IoT, this need has become more urgent, putting immense pressure on existing
network infrastructures. Processing ensures that data from IoT devices is converted into meaningful information
and actionable insights, which is crucial for real-time decision-making, optimization, and efficiency.

@VTUPADHAI – YT CHANNEL 2
 INTERNET OF THINGS (21CS735) MODULE 3 NOTES

Key considerations in IoT data processing:

 When to process?
 What to process?
These decisions depend on the type of data and the urgency of processing, which is categorized into three types
based on how quickly it needs to be processed:

Types of Data Processing in IoT

1. Very Time-Critical Data:
o Examples: Flight control systems, healthcare applications (e.g., life-support systems).
o Processing Latency: Must be processed within milliseconds.
o Requirements: Immediate decision-making is necessary. The data must be processed close
to the source (edge computing) to minimize latency.
2. Time-Critical Data:
o Examples: Vehicle systems, smart traffic management, smart homes, machine systems, and
surveillance.
o Processing Latency: Can tolerate a delay of a few seconds.
o Requirements: Data may be processed remotely (e.g., cloud computing), as the urgency is
lower than very time-critical data, allowing for collaborative or distributed processing.
3. Normal Data:
o Examples: Agricultural monitoring, environmental data, and other less sensitive
applications.
o Processing Latency: Tolerates delays ranging from minutes to hours.
o Requirements: This data does not require immediate attention and can be processed
leisurely, often using centralized servers or cloud solutions.

Processing Topologies in IoT

Processing topologies in IoT are essential for determining how data collected by sensor nodes is handled,
processed, and utilized. The correct selection of a processing topology can result in significant savings in
bandwidth, energy, and infrastructure costs, while ensuring that the latency requirements of the IoT application
are met. Processing topologies are broadly classified into on-site and off-site processing paradigms, with the latter
being further divided into remote processing and collaborative processing.

@VTUPADHAI – YT CHANNEL 3
 INTERNET OF THINGS (21CS735) MODULE 3 NOTES

1) On-site Processing Topology

In an on-site processing topology, data is processed at the source itself, meaning that the sensor node both collects
and processes the data locally. This type of topology is crucial in applications with very low latency tolerance,
where even slight delays can result in catastrophic consequences. For example, in healthcare or flight control
systems, on-site processing ensures immediate and real-time decision-making, as the data is processed where
it is generated, avoiding network transmission delays.

Figure 6.2 - Event Detection Using an On-site Processing Topology

The steps of on-site processing in this example are as follows:
1. Event Detection: A fire event is detected using a temperature sensor.
2. Sensor Node Activation: The temperature sensor is connected to a sensor node, which processes
the sensed data locally.
3. Processing and Alert Generation: The sensor node processes the fire event data and generates an
alert based on the predefined conditions (e.g., temperature threshold).
4. Further Actions (Optional): The sensor node has the option to forward the processed data to a
remote infrastructure for further analysis or storage.
The key advantage of on-site processing is the reduced latency, ensuring that critical events are handled
immediately without waiting for data to be transmitted across a network.

2) Off-site Processing Topology

In an off-site processing topology, the data is collected by the sensor node but processed elsewhere, typically at
a remote location such as a cloud server. This topology is suitable for applications that can tolerate some latency
and require less immediate data processing. Off-site processing is typically cheaper since the processing
infrastructure doesn't need to be on-site, allowing for more scalable solutions in large-scale IoT deployments.

@VTUPADHAI – YT CHANNEL 4
 INTERNET OF THINGS (21CS735) MODULE 3 NOTES

Remote Processing
Remote processing is the most common type of off-site processing, where data from multiple sensor nodes is
forwarded to a remote server or cloud infrastructure for processing. This allows for massive cost and energy
savings because powerful processing platforms can be shared among many smaller sensor nodes.

Figure 6.3 - Event Detection Using an Off-site Remote Processing Topology

The steps of remote processing are as follows:
1. Event Detection: A fire event is detected by a temperature sensor.
2. Sensor Node Sensing: The sensor node collects the event data but does not process it locally.
3. Data Transmission: The data is transmitted to a remote location (e.g., cloud or server) for further
processing and decision-making.
4. Processing and Decision Making: The remote infrastructure processes the data and makes the
necessary decisions or alerts.
The advantage of remote processing is that it allows for scalability and resource sharing among multiple nodes.
However, it relies heavily on network connectivity, and the network bandwidth must support the data
transmission.

Collaborative Processing
Collaborative processing is used in scenarios with limited or no network connectivity. In this topology, sensor
nodes collaborate by pooling their processing power to handle the data locally, without the need for a remote
processing infrastructure. This is useful in large-scale IoT deployments where network access may be intermittent
or unavailable, such as in agriculture or rural environments.

@VTUPADHAI – YT CHANNEL 5
 INTERNET OF THINGS (21CS735) MODULE 3 NOTES

Figure 6.4 - Event Detection Using a Collaborative Processing Topology

The steps of collaborative processing are as follows:
1. Event Detection: A fire event is detected by a temperature sensor.
2. Collaboration Among Nodes: Multiple nearby sensor nodes collaborate to process the data locally
without relying on a remote server.
3. Data Processing and Decision Making: The sensor nodes collaboratively process the data and
make the necessary decisions.
The key benefit of collaborative processing is its self-sufficiency and ability to reduce network traffic, which is
especially important in areas with limited connectivity. Additionally, collaborative processing can conserve
bandwidth and reduce latency compared to remote processing.

IoT Device Design and Selection Considerations

When designing an IoT solution, especially for sensor nodes, selecting the right processor is paramount. This
selection impacts various facets of the IoT system's performance, including usability, design, affordability, and
energy efficiency. Key considerations when selecting a processor for IoT devices include:
1. Size:
The size of the processor influences the form factor of the sensor node. A larger form factor typically
means higher energy consumption, which is not suitable for most IoT applications, especially those
requiring compact designs like wearables.
2. Energy:
Energy efficiency is crucial in IoT solutions. Processors with higher energy consumption necessitate
frequent battery replacements, reducing the sustainability of the IoT system. Hence, low-energy
processors are preferred for long-term, remote applications.
@VTUPADHAI – YT CHANNEL 6
 INTERNET OF THINGS (21CS735) MODULE 3 NOTES

3. Cost:
The cost of the processor and sensors affects the scalability of IoT solutions. Affordable processors
allow for higher-density deployments, making them more accessible for widespread applications
like gas and fire detection.
4. Memory:
Memory, both volatile and non-volatile, determines the device's capability to process, store, and
manage data. Devices with more memory offer advanced functionalities like local data processing
but are typically more expensive.
5. ProcessingPower:
The processing power of the device is essential, particularly for applications handling large data,
such as video or image processing. Devices with higher processing power support more complex
and data-intensive tasks.
6. I/ORating:
The input-output (I/O) voltage rating affects circuit complexity and energy usage. Modern
processors have a lower I/O voltage rating (3.3 V) compared to older ones (5 V), requiring
additional circuitry for compatibility with legacy systems.
7. Add-ons:
Processors that come with built-in features such as ADC units, clock circuits, USB and ethernet
connections, and wireless access capabilities offer greater usability. These add-ons simplify the
hardware development process and enhance the flexibility and robustness of the IoT solution.
In essence, selecting the right processor for an IoT application involves a balance between these factors to ensure
the solution is both cost-effective and capable of meeting the system's performance demands.

Processing Offloading in IoT

Processing offloading in IoT refers to shifting the computational load from resource-constrained devices (like
sensors) to more powerful, external resources such as edge devices, fog nodes, remote servers, or cloud
infrastructure. The goal is to keep IoT devices simple, energy-efficient, and low-cost while enabling them to
perform complex tasks by offloading heavy computation to external systems.
By doing so, IoT systems can achieve scalability, save energy, and minimize hardware costs while ensuring
efficient processing of massive data generated by IoT devices.

@VTUPADHAI – YT CHANNEL 7
 INTERNET OF THINGS (21CS735) MODULE 3 NOTES

Explanation of Figure 6.5 (Processing Layers in IoT)

Figure 6.5 represents a typical IoT deployment, showing how data is generated, processed, and transmitted across
different layers—from local sensing to cloud-based storage. Here's a breakdown of the steps:
1. Sensing Layer (On-site Processing):
o Sensors such as temperature sensors or cameras collect data from the environment (e.g., fire
detection or surveillance).
o In some cases, the processing happens locally, where the sensor is directly integrated with
a processor.
o Local processing is done on-site for critical or time-sensitive applications.
2. Edge Processing:
o Data from the sensors can be processed on nearby devices within the local network.
o The edge layer processes data closer to the source, reducing data transmission to cloud or
fog. This could involve aggregation, filtering, or manipulation of data.
o Communication is mostly via short-range wireless connections between local network
clusters.
3. Fog Processing:
o Fog nodes handle processing at an intermediate layer between local edge devices and the
cloud.

@VTUPADHAI – YT CHANNEL 8
 INTERNET OF THINGS (21CS735) MODULE 3 NOTES

o They offer faster response times than cloud processing because they are located closer to
the data source but still outside the immediate sensing layer.
o Fog processing helps reduce latency and network bandwidth usage.
4. Cloud Processing:
o If additional processing is required, the data is sent further to cloud servers via long-range
wireless connections (e.g., through a backbone network).
o Cloud processing involves large-scale, centralized systems capable of handling massive
amounts of data. However, it is slower due to network delays and costs related to bandwidth.

Three Key Parts of Data Offloading:

1. Offload Location:
o Edge: Processing is done close to the source of data generation, at the edge of the network.
For example, a smart camera performing object detection locally would be edge processing.
o Fog: Data is processed at fog nodes between the edge and the cloud. For instance, in smart
cities, fog nodes in the vicinity might handle traffic data processing, reducing the need for
cloud communication.
o Remote Server: Some IoT applications offload processing to remote servers. An example
would be a smart home system sending data to a dedicated server for processing commands
like turning lights on or off.
o Cloud: Data is sent to the cloud for more scalable processing. An example would be a
weather monitoring system sending environmental data to the cloud for extensive analysis
and prediction.
2. Offload Decision Making:
o Naive Approach: Involves offloading based on pre-defined rules, such as always sending
data to the nearest available resource. For example, temperature data from sensors might
always be sent to the closest fog node without considering network congestion.
o Bargaining-based Approach: Optimizes offloading by balancing trade-offs between
different parameters like bandwidth and latency. For example, using game theory, an IoT
system might allocate more resources to fire detection sensors (critical) while allocating
fewer resources to weather sensors (less critical).
o Learning-based Approach: Uses machine learning to make dynamic decisions based on
historical data. For example, a smart city IoT system could learn from past data to prioritize
certain types of sensors during specific times of day based on traffic patterns.
3. Offloading Considerations:

@VTUPADHAI – YT CHANNEL 9
 INTERNET OF THINGS (21CS735) MODULE 3 NOTES

o Bandwidth: If a network has limited bandwidth, offloading decisions may prioritize smaller
data volumes. For example, in a farm monitoring system, smaller sensor data might be
processed locally, while larger datasets are offloaded to the cloud.
o Latency: Critical tasks like fire detection must have minimal delay, so processing might
occur at the edge or fog level. Less critical tasks (like monitoring soil moisture) could
tolerate higher latencies and thus be processed in the cloud.
o Criticality: Highly critical tasks like surveillance in a secure facility will demand low-
latency, local processing. Less critical tasks like daily temperature monitoring could be
offloaded to remote servers or the cloud.
o Resources: Offloading decisions depend on the processing capacity of the destination. For
example, complex image processing may be offloaded to the cloud, while simpler
calculations (like averaging sensor readings) can be handled at the edge.
o Data Volume: Large volumes of data (such as video streams) are often offloaded to the
cloud, while smaller data (such as scalar sensor readings) may be processed locally or at the
edge. For instance, a smart surveillance system might offload video data to the cloud for
further analysis and storage.

@VTUPADHAI – YT CHANNEL 10
 INTERNET OF THINGS (21CS735) MODULE 3 NOTES

Question Bank-
(i) What are the different data formats found in IoT network tra c streams?
(ii) Depending on the urgency of data processing, how are IoT data classi ed?
(iii) Highlight the pros and cons of on-site and o-site processing.
(iv) Differentiate between structured and unstructured data.
(v) How is collaborative processing different from remote processing?
(vi) What are the critical factors to be considered during the design of IoT devices?
(vii) What are the typical data offload locations available in the context of IoT?
(viii) What are the various decision making approaches chosen for offloading data in IoT?
(ix) What factors are to be considered while deciding on the data offload location?

@VTUPADHAI – YT CHANNEL 11