IoT Exam Prep — 2026 Sommer

Goal

Pass by being able to choose, compare, calculate, and justify—not merely define terms. The supplied 2021 exam and question sheets repeatedly test scenario selection, protocol trade-offs, wildcard/QoS reasoning, and small calculations.

Scope map

1. IoT systems and architectures
2. IEEE 802.15.4
3. NB-IoT
4. 6LoWPAN
5. MQTT
6. IoT update protocols
7. IoT data models
8. Stream processing and complex events
9. Embedded AI

Highest-yield checklist

Before the exam, explain from memory:

• cloud vs edge vs mesh and when a gateway is needed;
• PHY vs MAC, FFD vs RFD, star vs peer-to-peer, 16-bit vs 64-bit addresses;
• unslotted vs slotted CSMA/CA, GTS, beacons, backoff, TDMA/FDMA, TSCH;
• NB-IoT features, three deployment modes, signaling/data channels, cell selection/reselection, tracking areas, TAU, eDRX/PSM;
• why IPv6 needs 6LoWPAN, fragmentation headers/offset units, stateless autoconfiguration, HC1/HC2 vs IPHC/NHC;
• MQTT topic matching, retain, Last Will, topic aliases, shared subscriptions, clean start/session expiry, and QoS;
• periodic/distance/dead-reckoning updates and the difference from stream compression;
• key-value vs linked data, continuous-query windows, stream processing vs CEP/ECA;
• edge/cloud AI placement, pruning, quantization, magnitude-pruning failure, and binarization.

Calculations to drill

• Periodic reporting: f_min = v_max / acc, so interval_max = acc / v_max.
• Distance reporting: report when dist(current, last_reported) >= acc.
• Dead reckoning: send a correction when prediction error exceeds the allowed threshold.
• 6LoWPAN offset: offset = start_byte / 8; with 72-byte payloads, fragment 5 starts at 288 bytes, so offset 36.
• MQTT delivery QoS: outgoing QoS = min(publication QoS, subscription QoS).

Answer audit: fix these before memorizing

The supplied answers are useful drafts, but these points are likely to cost marks:

Area	Correction
802.15.4 CSMA/CA	Explain the random draw: a doubled window can still produce a smaller actual backoff. Also say that larger windows reduce collision probability; do not claim every delay becomes longer.
802.15.4 topology	Large area does not itself select slotted CSMA/CA and 802.15.4 does not define multi-hop routing. Justify slotted operation with synchronization/sleep/GTS needs.
802.15.4 comparison	CSMA/CA controls access in time; frequency hopping changes carrier frequency. They are not competing definitions of the same mechanism.
NB-IoT channels	Paging is sent on a signaling/control channel, not a data channel.
6LoWPAN fragmentation	It is below IPv6 and enables the 802.15.4 link; it does not eliminate IP fragmentation entirely.
6LoWPAN header compression	HC1/HC2 are strongest for inferable link-local/UDP fields; global, multicast, routed, or non-default fields may need more bytes.
MQTT topic filters	+/+/temperature does not match the two-level topic basement/temperature; use # for all three supplied topics.
MQTT topic wildcards	A publish topic cannot contain + or #; home/+/temp is not a valid publish topic.
MQTT bathroom example	bathroom/+/temp matches exactly one middle level. It does not match bathroom/boiler/outlet/temp, which has two levels; use a consistent hierarchy or a broader filter plus application filtering.
MQTT topic alias	The broker stores the per-connection alias mapping so it can resolve and route later alias-only messages. It is not primarily about deciding who subscribed.
Update protocols	Update protocols reduce the number of updates; stream compression reduces the size of encoded data. Prefer distance-based/dead reckoning as the lecture-specific alternative for moving-location examples.
Stream vs CEP	Stream processing is database/DSMS-style query/aggregation over timestamped data streams; CEP is publish/subscribe-style event filtering/composition and pattern detection.
ECA rules	Do not call ECA always stateless: the lecture permits temporal, spatial, aggregation, and historical conditions.

Practice sequence

Pass 1 — build recall

Read one concept page, close it, and write the comparison table or process from memory. Mark any missing item as learning; do not reread passively.

Pass 2 — timed questions

Use the question bank and the original sheets. For each scenario answer in this order:

1. name the protocol/architecture;
2. state the relevant constraint;
3. compare the rejected alternative;
4. mention one limitation or trade-off;
5. calculate if applicable.

Pass 3 — exam simulation

Do a closed-book, timed run using the 2021 exam. Reserve time to check units, fragment offsets, MQTT topic depth, QoS direction, and whether the question asks for a justification. Reattempt every missed question the next day without looking at the model answer.

Suggested study rhythm (no exam date supplied)

• Cycle A: systems + 802.15.4; draw the stacks and CSMA variants.
• Cycle B: NB-IoT + 6LoWPAN; rehearse cell tracking and fragment-offset calculations.
• Cycle C: MQTT; solve topic filters and QoS/retain/shared-subscription cases by hand.
• Cycle D: update protocols + data models + streams/CEP.
• Cycle E: embedded AI + full timed paper.
• Repeat only weak cycles; aim for two consecutive closed-book runs at a safe margin rather than one memorized pass.

Source inventory

• Lectures: master-cs/raw/iot/1-Introduction.md, 2-IoT-Systems.md, 3.1-IoT_Internetworking-Motivation-802.15.4.md, 3.2-IoT_Internetworking-NB-IoT.md, 3.3-IoT_Internetworking-6LoWPAN.md, 3.4-IoT_Internetworking-MQTT.md, 4.1-IoT-Connected Data - UpdateProtocols.md, 4.2-IoT-Connected Data - DataModels.md, 4.3-IoT-Connected Data - Streams and Complex Events.md, 4.6-IoT-Connected Data - Machine Learning.md.
• Past exam: master-cs/raw/iot/2021 alte klausur.md.
• Practice sheets: master-cs/raw/iot/Question Sheet Week 1.md through Question Sheet Week 6.md.

Current weak areas

• CSMA/CA vs frequency hopping and the scenario justification.
• NB-IoT channel types and paging.
• 6LoWPAN fragmentation vs IP fragmentation.
• MQTT wildcard depth and invalid publish wildcards.
• Update protocol vs stream compression.
• Continuous queries/windows vs CEP/ECA.
• Embedded-AI trade-offs and pruning/quantization.

Update this checklist after each quiz attempt.