Cloud, Web & Mobile — Exam Prep (2026 Winter)

Status: stub, growing weekly. Only lectures 1–2 are ingested (week 2 of the term). Update after every lecture. Open items: exam date, exam format (written/oral), allowed aids, weighting. Fill in as soon as announced.

Goal

Be able to classify, compute, and justify at data-center scale — not just define terms. The material so far rewards: picking the right service model for a scenario, availability arithmetic, buy-vs-rent reasoning, failure-domain analysis, and cost model selection.

Scope map

Lecture	Topic	Concept page	Confidence
1	Historie, Motivation, Begrifflichkeiten (SaaS/PaaS/IaaS/DaaS, SLAs, billing)	cloud-history-and-terminology	new
2	Rechenzentren & IaaS (failure, replication, security, instances, VPC)	data-centers-and-iaas	new
3+	not yet ingested	—	—

High-risk items

Ranked by “easy to lose points and easy to fix”:

1. Metcalfe vs. Reed — quadratic pairs n(n−1)/2 vs. exponential subgroups 2ⁿ. Mismatching them is a classic point-loss.
2. Two-sided vs. same-sided network effect — the developer/user platform is the two-sided example, telephone/social networks are same-sided.
3. Availability arithmetic — 99.9% ⇒ 8.76 h/year, 99.99% ⇒ ~53 min, 99.95% ⇒ ~4.4 h. Do the conversion from the percentage and state the unit.
4. Buy-vs-rent crossover — rent_per_period × n < buy_price; then name the strategy (offline = decide up front, online = switch later) and the bounded-loss argument (worst case ≈ 2× the offline optimum).
5. Failure domains — order them: server < rack (switch, network cable, power cable, cooling) < cluster < data center. Same-rack replication is the worst case, not the cheap one.
6. Replication ≠ backup — one-line trap question.
7. Spot / reserved / on-demand — reason from the load profile: baseline ⇒ reserved, spiky ⇒ on-demand, interruptible ⇒ spot.
8. Elastic IP — the point is avoiding DNS propagation delay, not “static IP”; the provider reroutes internally.
9. DMZ — only DMZ hosts are reachable from the Internet; everything else is behind the router.
10. “Software must treat failure as the normal case” — the argument that justifies cheap standard hardware; pair it with persisting intermediate results and replication.

Study plan (rolling)

When	Task	Page
After lecture 3	ingest the new source, extend the MOC and this page	concepts/cloud-web-mobile/_index.md
Weekly	10 min: re-derive availability hours + one buy-vs-rent crossover from memory	concept page §8, §4
Before mid-term	self-quiz: 5 scenario questions (“which *aaS is this?“) + 3 failure-domain questions	write them into wiki/quizzes/
Week before exam	full mock in the format announced (once known)	create 2026-winter-cloud-web-mobile-mock-1.md

Weak concepts / re-test queue

Item	Why	Next action
—	nothing quizzed yet	first quiz after lecture 3

Log

• 2026-10-19 — page created; lectures 1–2 ingested (week 2 of term). No quiz taken yet.