IoT Value and Product Strategy · Study deck

IoT Systems Evolution: Computing and Placement

Computing improved, but a sensor system still cannot send every byte to an unlimited cloud without cost, delay, or energy consequences.

Ada is your guide for this deck.

systemsevolution
Ada, the module guide, in a scene from this chapter.
iotclass.org

After studying this chapter

Learning objectives

You will be able to:

  • Explain: Companies that rip out centralized servers in favor of edge-only architectures discover they still need cloud for ML training, fleet management, and long-term analytics.
  • Explain: Misconception 2: "Moore's Law ended, so computing improvement stopped.": Long-run transistor-density gains continued after real technologies departed from ideal Dennard scaling.
  • Explain: Present the “10× cycle” as a teaching heuristic only; avoid saying each era had 1,000× more devices or 1/100th cost.
  • Explain: Local intelligence is an architecture choice used when latency, resilience, privacy, bandwidth or safety justifies processing near the physical process.
iotclass.org

Major section

Distributed Computing Economics

Today's $0.50 microcontroller has more computing power than a $10,000 desktop computer from 1995.

  • Edge systems perform local analysis, fog nodes coordinate a site or region, and cloud platforms keep long histories or train complex models.
  • Raw data should shrink as it moves upward while the value and scope of insight increase.
  • Placement follows response time, bandwidth, privacy, resilience, and coordination needs rather than a fixed edge-first rule.

Why it matters

AI at the Edge: On-device camera inference can reduce upstream video traffic substantially; quantify bandwidth reduction for a defined resolution, frame rate and upload policy.

Fog computing hierarchy showing where device, edge, fog, and cloud layers process data according to latency and coordination needs.
Fog computing hierarchy showing where device, edge, fog, and cloud layers process data according to latency and coordination needs.
iotclass.org

Major section

IoT Systems Evolution Myths

Misconception 1: "IoT is just about connecting things to the internet.": Simple connected sensors are valid IoT endpoints.

  • Local intelligence is an architecture choice used when latency, resilience, privacy, bandwidth or safety justifies processing near the physical process.
  • Misconception 2: "Moore's Law ended, so computing improvement stopped.": Long-run transistor-density gains continued after real technologies departed from ideal Dennard scaling.
  • Misconception 3: "Cloud computing and IoT are competing approaches.": Cloud and edge are complementary, not competing.
  • Misconception 4: "IoT just needed cheaper hardware to happen.": Lower component cost was one contributor, not a prerequisite acting alone.
iotclass.org

Major section

Systems Evolution Pitfalls

Companies that rip out centralized servers in favor of edge-only architectures discover they still need cloud for ML training, fleet management, and long-term analytics.

  • The correct approach is a complementary multi-tier architecture, not wholesale replacement.
  • Pitfall 2: Extrapolating Moore's Law linearly into device cost: A $0.50 microcontroller does not mean a $0.50 IoT device.
  • They require sub-$1 hardware, 10+ year battery life, and operation in harsh environments (underwater, underground, extreme temperatures).
iotclass.org

Major section

Computing Power Revolution

MCU use also changes from about 100 mW typical to 50 mW active and 5 mW asleep.

  • Lower entry cost widens feasible deployments; it does not remove lifecycle, network, or cloud operating cost.
Systems evolution economics: falling hardware, connectivity, and cloud costs lowered the adoption threshold for IoT.
Systems evolution economics: falling hardware, connectivity, and cloud costs lowered the adoption threshold for IoT.
iotclass.org

Major section

Summary

Present the “10× cycle” as a teaching heuristic only; avoid saying each era had 1,000× more devices or 1/100th cost.

  • The mid-2000s power-efficiency pivot was one contributor to greater emphasis on efficiency, multicore processors and specialised logic.
  • Integration and component-cost trends expanded the capability available in microcontrollers, radios and sensors, but complete-node and deployment costs still matter.
  • Integration, efficiency, radios, networks and software made connected computing practical for many more physical processes.
Multi-factor convergence widened practical options for edge computing, fog computing, edge AI and wireless sensor networks.
Multi-factor convergence widened practical options for edge computing, fog computing, edge AI and wireless sensor networks.
iotclass.org

Deck summary

Key takeaways

Today's $0.50 microcontroller has more computing power than a $10,000 desktop computer from 1995.

  • Misconception 1: "IoT is just about connecting things to the internet.": Simple connected sensors are valid IoT endpoints.
  • Companies that rip out centralized servers in favor of edge-only architectures discover they still need cloud for ML training, fleet management, and long-term analytics.
  • MCU use also changes from about 100 mW typical to 50 mW active and 5 mW asleep.
  • Present the “10× cycle” as a teaching heuristic only; avoid saying each era had 1,000× more devices or 1/100th cost.
iotclass.org

Retrieval practice

Recall check 1 of 5

Ada says: answer from memory, then check your reasoning.

Q1A factory needs a safety-critical machine shutoff, site-wide coordination, and long-term fleet learning. Where should each job run?

ARun machine shutoff in cloud so it waits for remote decisions
BShutoff runs locally, site work at fog and fleet learning in cloud.
CAll jobs on each sensor so every node stores fleet history
DShutoff in cloud, fleet learning at the edge, site work nowhere
Show answer

Answer: B Correct: keep the safety action in the local controller; fog coordinates the site and cloud uses fleet history.

iotclass.org

Retrieval practice

Recall check 2 of 5

Ada says: answer from memory, then check your reasoning.

Q2Which challenge contributed to the departure from ideal Dennard scaling in the early-to-mid-2000s?

AGlobal chip manufacturing capacity was reached
BTransistors became too expensive to manufacture
CSoftware couldn't keep up with hardware improvements
DTransistors became so small that leakage current created excessive heat
Show answer

Answer: D Ideal Dennard scaling assumed smaller transistors could run at lower voltage without higher power density.

iotclass.org

Retrieval practice

Recall check 3 of 5

Ada says: answer from memory, then check your reasoning.

Q3Why were large-scale distributed IoT deployments less common in the 1990s than today?

ASensors, compute, radios, networks, platforms and deployment economics had not yet converged at modern scale
BGovernment restrictions kept civilian sensor fleets small across many markets
CConsumer demand for connected products remained too limited to sustain large fleets
DWireless networking standards arrived too late to support commercial fleets
Show answer

Answer: A Earlier connected systems were real.

Q4Complete the healthcare IoT vital signs aggregator:

Awindow = readings[-window_size:]
Bwindow = readings[:window_size]
Cwindow = readings[window_size:]
Dwindow = readings[::window_size]
Show answer

Answer: A Negative slicing [-window_size:] takes the last N elements for a moving average.

iotclass.org

Retrieval practice

Recall check 4 of 5

Ada says: answer from memory, then check your reasoning.

Q5A farm is evaluating a large soil-sensor deployment. Which economic change most directly improves feasibility as the node count grows?

ABroadband finally had enough capacity for small IoT telemetry streams
BFalling device costs made everyday sensing viable at scale
CWi-Fi and Bluetooth maturity after 2010 drove sensor affordability
DCloud platforms only became available after 2010 for device analytics
Show answer

Answer: B Falling total node cost improves the economics of dense sensing, but cost is not the only factor.

iotclass.org

Retrieval practice

Recall check 5 of 5

Ada says: answer from memory, then check your reasoning.

Q6Before accepting the chapter's illustrative claim that distributed factory placement saves $98K upfront, which evidence is most important?

ACheck network capacity for 1,000 devices as the cost proof.
BBOM, gateway, network, software, ops and workload data decide cost.
CCheck external sensor fit as the factory cost proof.
DOnly the historical price of a microcontroller
Show answer

Answer: B Treat the $98K result as an illustrative scenario, not a universal historical fact.

iotclass.org

Print reference

Answers 1 of 2

Answer key.

  1. B · Correct: keep the safety action in the local controller; fog coordinates the site and cloud uses fleet history.
  2. D · Ideal Dennard scaling assumed smaller transistors could run at lower voltage without higher power density.
  3. A · Earlier connected systems were real.
  4. A · Negative slicing [-window_size:] takes the last N elements for a moving average.
  5. B · Falling total node cost improves the economics of dense sensing, but cost is not the only factor.
iotclass.org

Print reference

Answers 2 of 2

Answer key.

  1. B · Treat the $98K result as an illustrative scenario, not a universal historical fact.
iotclass.org