Faculty of Computing & Meta-Technology

DIU2013 Internet of Things (IOT)

Chapter 2: Potential of Internet of Things (IOT) and Its Architecture and

Infrastructure

Copyright © 2024 Universiti Pendidikan Sultan Idris

 Topics Covered
Lecture 2: Potential of Internet of Things (IOT) and Its Architecture and
Infrastructure
• Understand the potential case studies of IOT
• Understand the architecture and infrastructure of IOT
Potential Scenarios of
IOT Infrastructure
Potential Scenarios of IOT
Infrastructure
 Mini Quiz
• Any differences between the previous 2 images?
• Which one is better or are they equal?
• Any business consideration? What do your C-level managers say?
• Do you need to share data?
• Is it allowed to share data? (By the company or by the government?)
• Different opinions: CEO wanna share data, but the CTO doesn’t want it
 Hints
• Do you need all the use cases to talk to each other?
• Do you need all the use cases to subscribe data from each other?
• Give example scenario where sharing platform is plausible:
• When all use cases are owned by the same company
• When there’s agreement to sell data to 3rd party – may need to anonymize
the data
• When there are no private/personal data used
• When there is Intellectual Property Protection: Businesses sharing a
workspace establish guidelines for protecting confidential information,
trade secrets
IOT Architecture
 IOT Architecture: What would
happen in one transaction?
1. Microservice based architecture a.k.a. REST API based services
a.k.a. stateless communication
2. Sensor sends message using MQTT to IOT platform
3. The message arrives at Message Broker that checks that the
message is authorized and has valid authentication creds using the
Broker Auth Service
4. Message Broker transforms the message to AMQP and sends it to
Event Consumer Service
5. Event Consumer Service queries device spec from Device Admin
Service and transforms it into a universal format, then sends it
to Scene Evaluation Engine
 IOT Architecture: What would
happen?
6. Scene Evaluation Engine retrieves the rules associated with this
message type. If an external API needs to be called, that request would
be sent to External Services Module (which would interact with that
external API)
7. Scene Evaluation Engine will also evaluate whether it needs to send
a message to the same or another device within the IoT network. To do
that, an outbound message (action request) is sent to the Device
Interaction Engine
8. Device Interaction Engine then requests the device-specific info
from Device Admin Service and customizes the outward message
based on it
 IOT Architecture: What would
happen?
9. Device Interaction Engine sends the message to the Message
Broker, which then sends the message to the device over the network
PS:
• For certain devices, Protocol Converter Service may be needed to
convert from MQTT to and from the device-specific protocol that the
device uses
• For rules that are triggered by scheduling rather than when an
inbound message arrives, Schedule Management Service will execute
the rule
 Mini Quiz
1. What are the available message broker software out there?
2. Which one do you think is the best?
3. Compare in terms of specs e.g. max transactions per second,
capacity, pricing, ease of configuration, etc
4. Compile in a comparison matrix and submit
here: https://forms.gle/4QbjxhuUjyUmqKNK8
IOT Architecture: Microservices
 Mini Quiz:
1. Anything wrong with the previous architecture?
 Potential Problem & Possible
Improvement
1. Take a look at the 'REST' all over the place. Most microservices are
RESTful. What's the problem?
Risk of blocking implementation; for every HTTP request, a thread is
assigned & the thread is held for the whole lifecycle of the request
a.k.a. limiting the number of concurrent requests (but not a
problem if there are only 500–1000 concurrent users)
2. Possible solution:
• Use the Message Broker for communication between individual microservices
• Use non-blocking approach; subsequent calls to different classes/methods &
external systems to run in a separate thread, hence letting the current thread
to do other tasks or return to the container pool
IOT Business Architecture
 Mini Quiz:
1. Any idea about IOT bundling? Pricing per unit?
2. Who owns the data? Platform? Each tenant? Customer's consent?
Country's regulation?
3. Can the data be "monetized"? Any terms and conditions to any
party involved?
IOT Platform Software Stack: Open
Source-driven
 So what's up w/ open-source
driven?
1. Easy to modify
2. Scalable
3. Upgradeable
4. Future proof (no vendor lock-in)
IOT Container Deployment on AWS:
Max Capacity 10k TPS
 How much does IOT platform cost?
1. Devices (preferably support MQTT & JSON):
• Device price varies subject to type
• Integration/installation fee was found about USD50k (may be waived for large deployment)
2. Estimated input load: 5,000 TPS from all tenants in 6 countries. Rule of thumb: roughly
200 TPS per less than 7 millions population (Ref: Managing & Assessing NaaS Capacity for
Telecommunication Sector on medium.com - also written by me)
3. 5,000 TPS costs USD50k/year on AWS (incl. hardware, storage, & analytics cluster)
4. New IOT solution varies subject to complexity, no of devices, coverage area, etc
5. New analytics capability costs could be similar to existing i.e. 2-6 weeks of dev
6. New GUI app may cost 8-12 weeks of dev
7. Connection cost i.e. Wired, 5G, Sigfox, satellite. Quite vary
8. New solution's scenes: subject to complexity design, but implementation could be less
than 5 mins
9. Overall time cost for new solution: 3-4 months (including testing)
 Cost Minimization possible?
1. Dockerization on AWS:
• It uses resources efficiently e.g. no duplicating of OS functions, thus...
• It pays only for what it uses
2. REST-dominated comm is quite burdening (HTTP-based), however still
capable (10k TPS were successfully tested) - But may replace it with a more
scalable message broker (is RabbitMQ the best?)
3. Smart anticipation: Build a generic solution blueprint to use as skeleton
THANK YOU!