Chapters

15 IoT Evolution: Systems Foundations

applications
iot
systems
evolution
Listen
Uses your device's voice. Tap ▶ beside any paragraph to hear just that one.
  • Podcast audio coming soon
  • Character narration audio coming soon
Lecture
Video lecture coming soon. The slides are the lecture for now.
Ask AI
Pick a prompt. We copy it for you and open an AI service; paste it there. Nothing is sent to IoTClass.

15.1 From a Local Switch to a Connected System

Connected systems change when cost, reach, and computation cross a useful threshold. This chapter traces those enabling constraints and separates evidence of adoption from claims about technology. The practical question is where useful computation belongs, and what makes that placement affordable to operate.

Begin with the farm pump. A local switch saved the farmer a walk. A timer saved routine work. A moisture sensor supplied evidence and reduced waste, and a small controller linked that evidence to action. A local display showed state; a nearby phone made control easier. A remote record then added history, while a shared service could compare many pumps. History helped find drift, and the wider view helped plan care. Each step changed both the value and the system boundary.

Responsibility expanded with that boundary. The farmer owned the switch, a maker owned some software, a network firm owned part of the path, and a service team owned long-term support. A bad switch affected one visit; a bad timer affected one schedule. A bad remote change could affect many sites. Wider reach made clear handoffs and review essential.

The outcome remained physical. A dry field still harmed crops, a dead battery still stopped a reading, and a broken valve still blocked water. Online success could not hide those failures. When adding a capability, name its value, owner, and new failure mode. Keep a safe local state, test the full path, and record the result.

The problem is choosing which capabilities help enough to justify their operating burden. Evolution offers choices rather than a fixed ladder: some jobs need one box, some need a shared service, and some need both. We will first examine the economics and power limits that created those choices, then assign work to devices, gateways, remote services, and people.

KEY CONCEPTS

  • Computing economics

    Lower unit cost creates options; lifecycle cost decides whether they last.

  • Power limits

    More transistors do not guarantee lower battery drain or less heat.

  • Compute placement

    Assign each task to the layer that meets its physical constraint.

  • State and authority

    Distributed copies need clear ownership when their values disagree.

15.2 Part Objectives

  • Trace IoT system changes through enabling constraints.
  • Separate adoption evidence from technology claims.

15.3 When Smaller Chips Change the Economics

The pump’s controller became practical through several changes working together. Chips became smaller and cheaper, and used less power for a given job. Memory and storage grew more affordable, wireless links became common, and online software platforms could support many products at once. No single change created modern IoT. The IoT Introduction uses the Five Verbs to describe what a connected system does; here we ask what made distributing those functions economically plausible.

Across computing eras, an expensive mainframe served an entire organization. PCs became affordable enough for businesses and homes to own multiple computers. Mobile devices put powerful connected computation in individuals’ pockets. Low-cost IoT chips can now disappear into products, infrastructure, and physical spaces. Computing became cheap enough, efficient enough, and connected enough to distribute intelligence across thousands of endpoints instead of concentrating everything in one central server.

The corresponding placement checklist is concrete. A 1960s business organized applications around scheduled access to a shared central machine. A 1980s office could justify putting a PC on a desk. A 2010s product could assume a phone, battery, camera, GPS, and network in a user’s pocket. A 2020s IoT sensor design asks whether a low-cost microcontroller, radio, and sensor can sit beside the physical process and remove delay, manual inspection, or unnecessary network traffic.

A microcontroller, or MCU, supplies local computation in such an endpoint. Once a capable wireless MCU costs only a few dollars, adding sensing, local logic, and connectivity to everyday objects becomes a deployment decision rather than a custom engineering exception. Smaller packaging alone does not explain this change: cost, efficiency, and connectivity have to work together.

The “10× technology cycle” is a classroom pattern for comparing orders of magnitude, not a measured historical law or a market forecast. Numerical comparisons need a named 1960s mainframe, 1980s PC, 2010s smartphone, and 2020s sensor model, each with a dated price and comparable unit-count source. Without those sources, the era comparison supports a question about affordability rather than an exact numerical claim. The lessons in IoT History and Paradigm Shifts help keep that distinction visible.

An economic crossover occurs when many low-cost edge nodes can serve a workload better than a central-only architecture on latency, bandwidth, or resilience. Cheap local computation gives the architect another option; it does not mean every sensor replaces a server. Use the pattern to ask which device categories may become viable next, then test the actual workload and operating cost.

15.4 Why More Transistors Do Not Remove the Power Limit

Affordability explains why a controller can sit beside the pump. It does not establish how long that controller can run from a battery. Moore’s Law describes growth in transistor density over generations. Ideal Dennard scaling described how shrinking transistors while lowering voltage could keep power density roughly steady. These are different claims: more transistors do not guarantee a proportionally faster or lower-power IoT device.

An important mid-2000s pivot came as real technologies departed from ideal Dennard scaling. Across multiple process generations, slower voltage scaling, leakage, interconnect, power-density, and thermal limits reduced reliance on frequency growth alone. Industry responses included multicore processors, specialised logic, and greater emphasis on performance per watt. Continued integration, lower-cost MCUs and radios, networking, and software-platform advances together widened the cases where distributed sensing and edge processing could outperform central-only designs.

The room heater provides another view of the same limits. A dial once set a target and one local box made the choice. Cheaper sensors and memory let that box track time, retain simple history, and show faults. Wireless links let a phone read state, support staff see some faults, and a user change a setting away from home. Each link also introduced another failure path. Fast safety work still needed a local response when the distant connection failed.

Neither integration nor connectivity removes physics. Small devices still have finite batteries, radios still consume energy, and heat still constrains chips. A cheaper component still needs a working enclosure. Measure the workload’s energy and heat rather than inferring them from a transistor trend. The distinction matters directly to whether the pump’s sensor can sleep between readings and remain serviceable over its intended life.

At 2.4 GHz, the 12.5 cm wavelength gives a 3.12 cm quarter-wave antenna that fits on a small board. That size does not establish battery life. A 220 mAh cell, after 1% annual self-discharge for ten years and 20% derating, leaves 159 mAh, or only 1.82 microamps average. After a 1.0 microamp sleep floor, roughly 0.82 microamps remain for waking and radio work—about 70.6 short reports per day under the linked lesson’s reporting assumptions.

The budget connects the physical antenna to the operating duty cycle: radio energy remains a system cost even when the processor is inexpensive. The guided Math Bridge works through wavelength, retention, sleep current, and wake time separately.

The maths

Battery Bruno's Math Bridge: work through the antenna and ten-year current budget.

Learn the maths

15.5 Give Each Layer a Job

With the cost and power limits clear, return to the pump’s work. Compute placement means choosing where each part of a computation runs: device, gateway, edge server, cloud service, or human workflow. The question is which layer suits each task. Device Evolution distinguishes embedded, connected, and IoT products; adding a connection does not by itself determine that division of work.

A soil sensor may sample and sleep locally, a LoRaWAN gateway may buffer readings, a cloud service may compare fields over a season, and an agronomist may approve irrigation changes. Cheap endpoint computation creates more placement choices while cloud coordination and human workflow remain part of the architecture. The technology-cycle figure in Part 2 connects falling unit cost and rising device scale to computation moving closer to physical processes.

At the device layer, assign sensing, actuation, sampling, filtering, low-power sleep, and immediate safety behavior. A gateway or edge layer can handle protocol translation, buffering, local rules, site-level dashboards, and operation during wide-area network (WAN) outages. Cloud and fleet services can handle long-term storage, model training, device registries, over-the-air (OTA) rollout, alert routing, and cross-site analytics.

The same division appears in industry: a vibration sensor may detect a threshold locally, a gateway aggregate samples, a cloud service train a model, and a maintenance planner decide whether to stop a line. Industrial IoT Fundamentals develops the Industrial IoT and Industry 4.0 context for these responsibilities.

A family’s watch, phone, car, lamp, and heating control show how widely computation has spread beyond one large machine. Some work happens inside each device, some nearby, and some in a large remote system. Put urgent work close to the event and long study where more history and power are available. Keep a safe local action when the distant path fails. The shift is from occasional data processing toward continuous sensing, local decisions, distributed services, and connected feedback loops.

Check knowledge

Check computing economics, enabling technologies, power limits, and placement choices.

Check systems foundations

15.6 Choose Placement by Constraint and Lifecycle Cost

The layer descriptions become useful only when tied to a constraint. Put computation near the device when response time, privacy, safety, radio airtime, or offline operation dominates. Cloud or server-side processing fits fleet history, large model training, cross-site comparison, heavy storage, and integration with enterprise systems. Edge Computing develops local patterns; Fog Computing examines the intermediate architecture.

Start with a concrete workload budget. A leak detector that wakes every few minutes, checks a threshold, and publishes a small MQTT or CoAP message may need only a low-power MCU and simple radio. A camera classifying frames locally may need an edge-AI module, a better power supply, heat management, and model-update storage. Edge AI/ML explores that machine-learning placement. A factory gateway aggregating 200 Modbus or OPC UA devices may need Linux, container isolation, local buffering, and a clear failover path to avoid losing events during a WAN outage.

A low-cost MCU is not a finished node. The bill of materials still includes sensor, radio, antenna, power regulation, enclosure, and mounting. Certification, provisioning, and device management add costs beyond those parts. ESP32, nRF52, STM32, RP2040, and ARM Cortex-M devices make many edge patterns affordable, but the system cost depends on battery life, installation labor, support, and replacement cycles.

Price the whole lifecycle. A cheaper endpoint can become expensive if it shortens battery life, requires truck rolls, cannot accept OTA updates, or needs manual certificate replacement. A slightly more expensive gateway may be cheaper over five years if it reduces cellular backhaul, keeps a line running during cloud outages, and gives operators useful diagnostics. Sale is only the start: storage, support, updates, field repair, and end-of-life work continue afterward.

For the farm pump, make that reasoning reviewable in three steps. First list the physical limits: latency, energy, bandwidth, privacy, safety, and environment. Then assign what runs in firmware, gateway, edge server, cloud service, or operator tool. Finally include provisioning, certificates, OTA updates, diagnostics, field replacement, and decommissioning in the lifecycle cost. Choose the placement that meets the physical constraint with the lowest credible lifecycle risk.

15.7 Keep Distributed State and Responsibility Consistent

Moving computation outward can create resilience, but it also creates more state. A local reading, gateway cache, MQTT retained message, cloud device twin, time-series row, and dashboard state may not update together. A device twin is a remote representation of device state. Define which layer is authoritative for commands, telemetry, alarms, and configuration when those representations disagree.

Cheap computation therefore needs protocol and operations discipline. BLE, Wi-Fi, Thread, LoRaWAN, LTE-M, MQTT, CoAP, OPC UA, and HTTP carry different assumptions about latency, power, addressing, reliability, and security. A sound design includes backpressure, buffering, idempotent commands, timestamp discipline, quality flags, least-privilege credentials, and firmware updates that can roll back. Idempotent commands avoid repeated effects when the same command is processed again; backpressure controls incoming work when a receiver cannot keep up.

Performance per watt is visible in operation. Low-power endpoint firmware spends much of its life sleeping, then wakes for interrupts, analog-to-digital converter (ADC) reads, radio joins, watchdog checks, and short publish windows. Gateways translating constrained field networks into IP services must preserve timestamps, units, quality codes, sequence numbers, and source identity. Cloud services receive derived state rather than perfect truth, so they need late-arrival handling, duplicate suppression, clock-skew tolerance, and explicit command expiry.

Security state is distributed too. A device may hold a hardware identity, a gateway enforce topic or route policy, a cloud registry own the device twin, and an operator console issue commands. Certificate rotation, revoked devices, OTA rollback, and factory resets cross these layers. More links create more entry points; more data creates more responsibility. Remote action raises the cost of a mistake, making clear ownership essential at every boundary.

Review three paths explicitly. The command path separates requested, queued, delivered, applied, rejected, overridden, and expired states. The telemetry path preserves timestamp, unit, calibration, and quality metadata before aggregation changes meaning. The operations path treats battery, signal strength, firmware version, queue depth, gateway health, and update status as first-class data. Reconnect behavior, failed-update recovery, queue backpressure, audit logs, and safe defaults keep the pump service operable when layers disagree.

15.8 Read Evolution as Evidence of New Choices

IoT has evolved through phases of increasing interconnectedness among devices, people, and systems, from simple networks toward integrated ecosystems. The useful milestone is a changed capability with an identifiable constraint and outcome. Real progress came unevenly and did not make earlier methods obsolete for every job. A system should use the smallest arrangement that proves its outcome.

The heater illustrates the operating consequence of scale. Shared software can help one team support many homes, discover common faults, and deliver updates across many products. A mistake can reach those products too. More devices create more failure points, and scale is useful only when the service can operate them. That is why the pump’s safe local state and the service’s ownership record remain necessary as its capabilities grow.

The IoT World Forum Barcelona video offers another view of interconnected systems. Watch it separately, carrying the chapter’s questions about cost, reach, placement, and responsibility into that account of evolution.

For a focused continuation of the architecture discussion, IoT Systems Evolution: Computing and Placement compares where tasks belong. Judge each proposed shift by the value it adds, the new failure it creates, and the evidence that the complete service still works.

Summary

Cheaper, smaller, more efficient chips, wireless links, and shared services widened the places where useful computation could live. The 10× cycle is a teaching pattern, while Moore’s Law and ideal Dennard scaling describe different relationships. The mid-2000s power-efficiency pivot increased the importance of performance per watt without eliminating radio, battery, or thermal limits.

The farm pump makes those changes concrete. Local sensing and safe action, nearby buffering, remote history, and human approval can each serve a different constraint. Evaluate their full lifecycle cost and define authority when distributed copies disagree. A working online path must still lead to evidence of the physical outcome.

Choose where each task runs by its physical constraint, then prove that the complete service remains affordable, safe, and operable as its reach grows.

Hi, I’m Ada. What do you need?

Find help

Let’s find what you need

Find a lesson or return to your module guide.

Give feedback

Help improve this chapter

Tell us what helped or what needs fixing.

Tips

Try a small next step

A few ways to get more from this chapter.

Ask AI

Continue the conversation

Take the conversation further with your preferred assistant.

AI help is available in supported modules when enabled in Settings.

No live person answers on independent study. Feedback does not promise a personal reply.

For welfare or safety concerns, follow your school’s procedures outside IoTClass.