Skip to main content

Courses About Pricing Educators

Sign in Get started

Courses About Pricing Educators

Get started Sign in

Loading courses

Loading lessons

Inside Distributed Systems: System Design: Uncharted

Courses System DesignInside Distributed Systems

Inside Distributed SystemsPro

Reason about drifting clocks, uncertain failure, quorums, consensus, leases, and coordination.

System Design
Unit 7 of 12
20 lessons
147 min
420 XP

This unit is part of Pro. Upgrade to open every lesson in it.

Lessons

20 total

1No One Has the TimeExplain why drifting wall clocks cannot establish causality between events on different machines.ProInteractiveConcept7 min20 XPOpen
2Happened BeforeIdentify send-and-receive relationships that establish happened-before order and distinguish them from concurrent events.ProInteractiveConcept7 min20 XPOpen
3Logical CountersUpdate a logical clock when messages arrive so causal order is preserved without claiming elapsed real time.ProInteractiveConcept7 min20 XPOpen
4The HeartbeatInterpret missing heartbeats as timeout-based suspicion rather than proof that a peer is dead.ProInteractiveConcept7 min20 XPOpen
5Dead or Slow?Choose a failure-detection timeout by balancing faster recovery against false suspicion of healthy but slow peers.ProInteractiveConcept7 min20 XPOpen
6Split BrainRecognize split brain when isolated leaders accept incompatible writes to the same logical state.ProInteractiveConcept7 min20 XPOpen
7Majority RulesDetermine which partition can form a majority and explain why two disjoint groups cannot both hold one in a fixed cluster.ProInteractiveConcept7 min20 XPOpen
8Napkin: R + W > NCalculate the minimum whole write quorum that guarantees read-write overlap for several replica configurations.ProInteractiveConcept7 min20 XPOpen
9Tune the QuorumChoose read and write quorum sizes that preserve overlap while placing more coordination cost on the less frequent operation.ProInteractiveConcept7 min20 XPOpen
10Electing a LeaderSequence a term-based leader election from follower timeout through majority vote and new leadership.ProInteractiveConcept7 min20 XPOpen
11The Stale LeaderUse monotonically increasing fencing tokens to make storage reject work from an expired leader or lock holder.ProInteractiveConcept7 min20 XPOpen
12Consensus, as a StoryPredict when a consensus group must stop committing so it preserves one safe log despite delay or partition.ProInteractiveConcept7 min20 XPOpen
13The Replicated LogExplain why replicas converge when they apply the same deterministic commands in one committed order.ProInteractiveConcept7 min20 XPOpen
14Distributed Locks Are SpicyExplain how a paused lease holder can resume stale work and why the downstream resource must enforce fencing order.ProInteractiveConcept7 min20 XPOpen
15LeasesExplain how expiring ownership restores progress after a holder disappears and what renewal and fencing must still protect.ProInteractiveConcept7 min20 XPOpen
16Who Is in the ClusterExplain how gossip membership trades immediate agreement for decentralized propagation and eventual convergence.ProInteractiveConcept7 min20 XPOpen
17The Coordination ServiceDesign a coordination seam where multiple application services share one service backed by consensus state.ProDiagramConcept10 min20 XPOpen
18Never Build This YourselfDecide when an existing transactional database is sufficient and when a proven coordination system is justified.ProInteractiveConcept7 min20 XPOpen
19Explain Split BrainExplain how a partition creates conflicting-leader risk and how majority quorums prevent dual commits at an availability cost.ProInteractiveConcept8 min20 XPOpen
20Distributed Systems Mastery CheckEvaluate distributed actions using causal order, majority progress rules, quorum overlap, and fencing-token freshness.ProInteractiveMastery check10 min40 XPOpen

Queues and Streams APIs and Realtime