34  Why Protocol Choice Is Hard

Competing Constraints, Layer Boundaries, and Evidence-First Shortlisting

fundamentals
protocol
framework
challenge

34.1 In 60 Seconds

IoT protocol selection is difficult because the strongest requirements usually compete. A radio that reaches farther often sends less data or needs more energy. A high-throughput link may need mains power or short range. A protocol that looks simple at the device may require gateway, service, update, and monitoring work. Start with hard constraints, separate protocol layers, and keep evidence for every eliminated option.

34.2 Start With the Story

Start with a product meeting where someone says just use the protocol we know, while the deployment quietly needs different range, power, latency, security, and operations evidence. The core idea in Why Protocol Choice Is Hard is simple: protocol choice is a constraint problem, not a popularity contest, and the defensible answer comes from eliminating bad fits before scoring good candidates. This page focuses that idea on Explains why IoT protocol choices are hard and how range, energy, latency, throughput, security, scale, interoperability, and operations shape shortlists. In everyday IoT, parking sensors, wearables, logistics tags, and building retrofits need different proof even when they all send small messages. Start simple: state the hard constraints, reject impossible options, compare the finalists, and then validate the chosen protocol in the field.

Phoebe the physics guide

Phoebe’s Why

“Range” in the constraint triangle is not one fixed number a protocol carries around – it is what is left over after a link budget, and gain is one of the few line items in that budget an installer can change without touching the radio chip at all. An antenna with gain over isotropic does not create power; it reshapes where the existing power goes, trading coverage angle for concentration along a preferred direction. That reshaping raises the effective isotropic radiated power the far end sees, which plugs straight into the same free-space path-loss relationship that sets range for every protocol in this chapter’s comparison table – LPWAN, cellular, Wi-Fi, or BLE alike.

The Derivation

Effective isotropic radiated power combines transmit power and antenna gain:

\[\mathrm{EIRP\,(dBm)} = P_t\,\mathrm{(dBm)} + G\,\mathrm{(dBi)}\]

Free-space path loss grows with distance and frequency:

\[\mathrm{FSPL\,(dB)} = 20\log_{10}d + 20\log_{10}f - 147.55 \quad (d\text{ in m}, f\text{ in Hz})\]

A link closes while received power clears the receiver’s sensitivity floor:

\[P_{rx} = \mathrm{EIRP} - \mathrm{FSPL} + G_{rx} \geq P_{sensitivity}\]

so the maximum tolerable path loss, and therefore the maximum range, rises exactly with any added gain:

\[\mathrm{FSPL}_{max} = \mathrm{EIRP} + G_{rx} - P_{sensitivity}\]

Worked Numbers: A LoRaWAN Uplink Budget

The chapter names LoRaWAN as a long-range LPWAN candidate but does not fix a link budget, so take standard values: a 14 dBm (25 mW) end node at 868 MHz, a \(-130\) dBm gateway sensitivity, and a 0 dBi node antenna.

  • With a 0 dBi gateway antenna: \(\mathrm{FSPL}_{max} = 14 + 0 - (-130) = 144\) dB, which solves to \(d \approx 435\) km of free-space ceiling
  • Swap in a 6 dBi gateway antenna: \(\mathrm{FSPL}_{max} = 14 + 6 - (-130) = 150\) dB, solving to \(d \approx 869\) km – a \(10^{6/20} = 1.995\times\) range increase from 6 dB of gain alone, matching the textbook “6 dB doubles range” rule almost exactly
  • Aperture view of the same result: \(A_e = G\lambda^2/4\pi\) grows from \(0.00951\ \mathrm{m}^2\) at 0 dBi to \(0.0378\ \mathrm{m}^2\) at 6 dBi – about 3.98x the collecting area, matching \(10^{6/10} = 3.98\)
  • These are free-space ceilings, not promised coverage – terrain, foliage, and Earth curvature cut real deployments far below either number, but the gain still doubles whatever the achievable range turns out to be

34.3 Protocol Selection: The Challenge

Protocol selection is not a popularity contest. It is a constraint-ranking problem. The hard part is that an IoT deployment asks for several things at once:

  • reach the device;
  • fit the energy source;
  • carry the payload at the required freshness;
  • respond within the latency budget;
  • scale to the device count;
  • protect the data and control path;
  • interoperate with the rest of the stack;
  • remain operable after installation.

No single protocol wins every dimension. The goal is to identify which constraint is non-negotiable, then find the smallest set of protocol candidates that can survive it.

34.4 Learning Objectives

By the end of this chapter, you will be able to:

  • Explain why range, energy, throughput, latency, coverage, security, scalability, interoperability, and operations compete.
  • Rank hard constraints before comparing protocol names.
  • Separate physical/link choices from network, transport, and application-layer choices.
  • Avoid layer-confusion questions such as “MQTT or Wi-Fi?”
  • Use a first-pass shortlist before doing weighted scoring or detailed lifecycle-cost work.
  • Record what evidence is still missing before a protocol can be selected.

Quick Check: Challenge Reasoning

Minimum Viable Understanding

Protocol selection starts with the dominant constraint. A camera, a wearable, a buried meter, an indoor light switch, and a mobile asset do not belong to the same connectivity class because their payload, range, power source, latency, and operations assumptions differ.

34.5 Why This Is Hard

Most failed protocol decisions begin with one of three shortcuts:

  • choosing a familiar protocol before checking power and coverage;
  • comparing protocol names that operate at different layers;
  • optimizing one dimension while ignoring the deployment system around it.

The challenge is not that the protocol landscape is random. The challenge is that every reasonable option is optimized for a different job.

Protocol-selection constraint map showing range, energy, throughput, latency, coverage, security, scalability, interoperability, and operations as competing review areas.
Protocol choice starts by ranking competing constraints before naming winners.

34.6 Selection Dimensions

Use these dimensions as a review checklist before naming a winner.

Dimension
Question
Evidence
Failure Mode
Range and coverage
Where are devices relative to gateways, access points, readers, phones, or towers?
Site survey, antenna plan, weak-path tests, mobility assumptions.
Devices install successfully but cannot report reliably.
Energy source
Can the device sleep, wake, transmit, receive, and recover within its energy budget?
Full reporting-cycle current trace and maintenance interval.
The device works in the lab but fails field lifetime targets.
Payload and freshness
How many bytes, how often, how bursty, and how fresh must the data be?
Payload budget, retry policy, acknowledgment needs, latency contract.
A high-throughput link is overbuilt, or a low-data-rate link misses timing needs.
Operations
Who provisions, updates, monitors, diagnoses, and replaces the system?
Provisioning workflow, update path, logging, support and replacement plan.
The radio works, but the deployed system is hard to maintain.

34.7 The Constraint Triangle

The common teaching shortcut is a triangle: range, energy, and throughput cannot all be optimized at once. The triangle is useful, but it should not be treated as a magic formula.

34.7.1 High Throughput

Useful for images, audio, firmware transfer, dense local backhaul, dashboards, and user-facing interactions. The trade-off is usually higher active energy, shorter practical range, or stronger infrastructure requirements.

34.7.2 Long Reach

Useful for sparse outdoor sensing, utility metering, agriculture, remote assets, and hard-to-wire sites. The trade-off is usually lower payload rate, stricter duty behavior, delayed downlink, or site-specific coverage validation.

34.7.3 Low Energy

Useful for devices that sleep most of the time and are expensive to service. The trade-off is usually smaller payloads, limited receive windows, less frequent reporting, or gateway/topology planning.

The safest reading is: the dominant constraint chooses the first candidate family. Secondary constraints then decide which member of that family survives.

Turn the triangle into measurements before it becomes a debate. A LoRaWAN field sensor needs received-signal margin, duty-cycle, downlink, and battery-life evidence for the real site. A BLE wearable needs phone-reconnect behavior, advertising interval, current trace, and privacy evidence. A Wi-Fi camera needs airtime, roaming, firmware-update, and mains-power assumptions. A Thread or Zigbee control network needs router density, commissioning success, route repair, channel planning, and local-failure behavior. The challenge answer is stronger when it names the test that would eliminate each tempting family.

34.8 Layer Boundaries

Many protocol arguments are actually layer mistakes. MQTT, CoAP, HTTP, Wi-Fi, BLE, LoRaWAN, IPv6, TCP, and UDP are not all alternatives to each other.

Detailed IoT protocol stack across five layers, each with representative protocols, data units, and example roles: application with CoAP, MQTT, HTTP/2, and AMQP; transport with UDP, TCP, DTLS, TLS, and QUIC; network with IPv6 addressing and routing; data link; and physical.
Layer-aware selection builds a stack instead of comparing unrelated technologies.

34.8.1 Layer Review

34.8.3 Network And Adaptation

Examples: IPv4, IPv6, 6LoWPAN, RPL, LoRaWAN network behavior.

This layer answers how packets are addressed, adapted, or routed.

34.8.4 Transport

Examples: TCP, UDP, QUIC, DTLS context.

This layer answers how endpoints exchange streams, datagrams, sessions, or protected transport.

34.8.5 Application

Examples: MQTT, CoAP, HTTP, AMQP, custom payload contracts.

This layer answers how application data is named, encoded, requested, published, or acknowledged.

Layer Mistake

“Should we use MQTT or Wi-Fi?” compares an application protocol with a link-layer technology. A complete stack might use both: Wi-Fi for local connectivity, IP and TCP for transport, and MQTT for application messaging.

34.9 First-Pass Shortlisting

A first-pass shortlist is not the final decision. It prevents impossible options from wasting design time.

Evidence-first protocol shortlist workflow: define the decision question, rank hard constraints, remove impossible families, then keep evidence for remaining candidates.
First-pass shortlisting removes impossible families before detailed scoring.

Use this sequence:

  1. Define the communication job. Is it telemetry, control, configuration, bulk transfer, discovery, local automation, tracking, or backhaul?
  2. Rank hard constraints. Identify what must not fail: power source, reach, latency, payload, security boundary, mobility, or operations.
  3. Eliminate impossible families. Remove options that cannot satisfy the hard constraint without redesigning the system.
  4. Keep a shortlist. Compare only the candidates that remain viable.
  5. Record missing evidence. Note field tests, current traces, payload measurements, standards profiles, or operations checks still required.

34.10 Example Challenge Reviews

These examples show the shape of the reasoning without pretending that one protocol is always correct.

34.10.1 Remote Telemetry

Dominant constraint: reach and energy.

Likely first-pass direction: LPWAN or cellular IoT candidates, depending on coverage evidence, payload freshness, operations, and ownership model.

Do not decide from: nominal maximum data rate alone.

34.10.2 Local Control

Dominant constraint: latency, reliability, and local coverage.

Likely first-pass direction: local star or mesh candidates, depending on power source, room geometry, gateway placement, and maintenance access.

Do not decide from: radio module price alone.

34.10.3 High-Data Device

Dominant constraint: throughput and user experience.

Likely first-pass direction: Wi-Fi, Ethernet, or cellular broadband candidates, depending on power source, mobility, coverage, and backhaul.

Do not decide from: sensor-node energy assumptions.

34.10.4 Moving Asset

Dominant constraint: coverage across changing locations.

Likely first-pass direction: carrier, gateway, or hybrid designs depending on routes, handoff, dead zones, and reporting criticality.

Do not decide from: one static site test.

34.11 Common Pitfalls

Pitfall 1: Asking One Protocol To Maximize Everything

If requirements demand long reach, high throughput, tiny energy use, low latency, low cost, and no infrastructure, the problem is under-ranked. The correct next step is constraint negotiation, not protocol shopping.

Pitfall 2: Comparing Layers Instead Of Building A Stack

“LoRaWAN or MQTT” and “Wi-Fi or CoAP” are layer mistakes. First identify the connectivity layer, then complete network, transport, security, and application decisions around it.

Pitfall 3: Treating Module Price As System Cost

The radio part is not the deployment. Gateways, subscriptions, antenna work, commissioning, monitoring, replacement, and field service can dominate the real decision.

Pitfall 4: Trusting Nominal Specs

Advertised range, data rate, and power numbers are starting points. Real sites need antenna placement, interference, payload, retries, receive behavior, and operations evidence.

34.12 Check Your Understanding

Knowledge Check: Dominant Constraint

Knowledge Check: Protocol Layers

Matching Check

Ordering Check

34.13 Place Each Candidate In Trade-Off Space

Protocol choice is hard because the things teams usually want, long range, high data rate, and low power, pull against one another. Radio physics ties them together: reaching farther for a given power usually means spending airtime on fewer bits through narrower bandwidth, lower coding rate, spreading, or more conservative modulation. Pushing many bits per second usually needs wider bandwidth, cleaner signal conditions, shorter practical range, stronger infrastructure, or more energy.

So every IoT radio is a point in a trade-off space. The honest question is not “which protocol is best?” but “which compromise fits this job?” A battery soil sensor that reports moisture every 30 minutes can tolerate tiny payloads and seconds of delay, so LoRaWAN or NB-IoT may be reasonable after coverage checks. A machine-vision camera on the same farm needs steady megabits per second and mains power, so Wi-Fi or Ethernet is more realistic. Same site, different communication job, different shortlist.

Technology Typical reach class Typical data-rate class Power posture
BLE Room to nearby-phone range About 1-2 Mbit/s at the PHY, less at the application Low when duty-cycled
Zigbee or IEEE 802.15.4 Room/building range, often extended by mesh 250 kbit/s PHY for 2.4 GHz 802.15.4 Low with careful sleep and routing design
Wi-Fi Local access-point range Tens to hundreds of Mbit/s depending on generation and link quality Higher active energy; often mains or rechargeable
LoRaWAN Long-range LPWAN class Small, infrequent payloads at low data rates Very low for sparse uplinks
NB-IoT Cellular wide-area class Small telemetry payloads, operator and coverage dependent Low power with cellular power-saving modes

Use the table as a placement aid, not as a winner list. Want long reach and years on a battery? LoRaWAN may fit, but the accepted cost is low throughput and constrained downlink behavior. Want to move a firmware image quickly? Wi-Fi may fit, but the accepted cost is access-point planning and higher active energy. Want a dense building control network? Thread or Zigbee may fit, but the accepted cost is commissioning, router density, route repair, and channel planning.

34.13.1 Hidden Coupling

Range, rate, and power are the headline axes, but real decisions add more: topology, spectrum regime, cost, scale, latency, mobility, ownership, provisioning, updates, monitoring, and failure recovery. A mesh can reduce gateway count, but it can increase route churn, commissioning work, and security-key operations. A cellular link can remove local gateways, but it adds SIM lifecycle, coverage contracts, operator outages, and recurring fees. A high-rate Wi-Fi design can simplify application updates while forcing larger batteries or mains power.

Two teams can therefore both need “wireless sensors” and still choose oppositely. A utility may choose NB-IoT for scattered meters because licensed coverage and no site-owned gateways matter more than subscription cost. A factory may choose Zigbee or Thread for dense in-building sensors because local control, no recurring carrier fee, and mesh coverage matter more than wide-area reach. Neither decision is automatically wrong; each weights the axes differently.

A defensible selection names the layer being chosen, the constraint that eliminated alternatives, and the evidence still missing. For example: “Thread remains a candidate for the building mesh, but stairwell packet delivery, commissioning success, and route repair after router loss must be measured.” That sentence is stronger than “Thread is modern” because it turns protocol preference into testable engineering work.

34.14 Summary

Protocol selection is difficult because useful requirements compete. The first task is not to choose the most familiar protocol; it is to rank hard constraints, understand which layer each technology belongs to, and remove options that cannot meet the deployment evidence. Once the shortlist is real, later chapters can compare candidates systematically.

34.15 See Also

34.16 What’s Next

34.16.1 Avoid Mistakes

Protocol Anti-Patterns shows what happens when teams skip the constraint checks in this chapter.

34.16.2 Apply A Process

Systematic Selection turns this challenge review into a repeatable elimination and scoring workflow.

34.16.3 Test Scenarios

Selection Scenarios applies the constraints to realistic deployment situations.

34.16.4 Use the Selector

Protocol Selector Wizard provides the wider shortlist helper for review.

34.17 Key Takeaway

A protocol recommendation should survive challenge. Defend it against latency, power, range, payload, reliability, security, cost, gateway, and lifecycle constraints.