IoT Systems and Architectures

Core idea

An IoT system is the complete hardware and software needed for a particular IoT application: smart objects, networking, gateways, online services, data management, and data processing. An IoT platform is reusable infrastructure and services used to build many such systems. Platforms may be horizontal (general-purpose, e.g. cloud provider platforms) or vertical (domain-specific, with specialized analytics).

IoT has two complementary views:

• Connected things / M2M: smart things cooperate using network and Internet technologies.
• Connected data / big data: things continuously produce data about the physical world, which is stored, queried, analyzed, and fed back into it.

The target is a digital twin: a global digital representation that is live, predictive, and interactive.

Architecture questions

Every IoT architecture must answer:

1. Which software blocks exist and how are they implemented?
2. Where are they executed: smart object, gateway/edge, or cloud?
3. How are they connected: which APIs and protocols?

The course simplifies the software stack to:

• IoT networking: connects all hardware and software parts.
• IoT data management: measures, stores, exposes, and queries data, often with distributed/NoSQL/graph databases.
• IoT data processing: filtering, visualization, analytics, prediction, and machine learning.

Placement patterns

Architecture	Placement	Strength	Cost/risk
Cloud-based	Devices are minimal; gateways mostly provide cross-cutting functions; processing and storage are in the cloud.	Simple devices and centralized management.	Network dependence, latency, bandwidth, privacy, cloud outage.
Edge-based	Gateways perform part of cloud data management/processing and may communicate with each other.	Lower latency, local processing, less traffic, better resilience.	More complex gateways and distributed coordination.
Mesh-based	Devices perform many functions and communicate directly; gateway/cloud are fallback or absent.	Local autonomy, resilience, and potentially low latency.	Powerful/energy-aware devices and harder distributed management.

Cloud platforms currently support cloud-based IoT widely; edge-based systems are increasingly important, while flexible mesh-based systems are the future direction in the lecture.

Gateways

A gateway is not always necessary. An IP-capable smart object can connect directly through an IP router. A gateway is useful when devices use a non-IP or incompatible protocol, need protocol translation, local buffering, security isolation, edge processing, or access to a remote cloud.

Two Internet-connection approaches:

• Application-layer proxy: devices use a dedicated stack (e.g. Zigbee/BLE); a gateway stores/translates data and exposes Internet protocols. It supports very constrained devices but adds a dependency and possible bottleneck.
• IP router: devices speak IP directly and are Internet hosts. This avoids an application proxy but requires more capable devices; constrained links may need 6LoWPAN.

Exam traps

• A platform is not the same thing as a complete application-specific system.
• Edge and mesh are not synonyms: edge moves cloud functions to gateways; mesh moves more functions into the devices and their peer network.
• Large physical area alone does not make a mesh: the 802.15.4 standard does not specify multi-hop routing; higher layers must provide it.

Sources

• master-cs/raw/iot/1-Introduction.md
• master-cs/raw/iot/2-IoT-Systems.md
• master-cs/raw/iot/Question Sheet Week 1.md