Skip to main content

Courses About Pricing Educators

Sign in Get started

Courses About Pricing Educators

Get started Sign in

Loading courses

Loading lessons

Scaling Out: System Design: Uncharted

Courses System DesignScaling Out

Scaling Out

Replace one overloaded machine with balanced, stateless, failure-aware service capacity.

System Design
Unit 2 of 12
20 lessons
157 min
420 XP

Every lesson is open. Sign in when you are ready to begin.

Sign in to start

Lessons

20 total

1The Lunch RushExplain why adding independent workers raises capacity only when incoming work can be distributed among them.InteractiveConcept7 min20 XPOpen
2Bigger Box or More BoxesChoose between scaling up and scaling out from capacity, failure-tolerance, and operational tradeoffs.InteractiveConcept7 min20 XPOpen
3The Ceiling of UpIdentify the finite capacity and single-failure boundary that vertical scaling cannot remove.InteractiveConcept7 min20 XPOpen
4Meet the Load BalancerModel a load balancer as the seam that distributes one audience across interchangeable service instances.DiagramConcept10 min20 XPOpen
5Balancing StrategiesSelect round robin, least connections, or stable hashing from the workload property each strategy serves.InteractiveConcept7 min20 XPOpen
6Health ChecksDescribe how health probes remove an unsafe instance from new traffic and qualify it to return.InteractiveConcept7 min20 XPOpen
7The Session ProblemDiagnose process-local session memory as the reason a rerouted user appears logged out.InteractiveConcept7 min20 XPOpen
8Sticky or SharedContrast sticky routing with shared sessions and justify when instance interchangeability matters more.InteractiveConcept7 min20 XPOpen
9Go StatelessConstruct a horizontally scaled request path whose services share session state instead of owning it locally.DiagramConcept10 min20 XPOpen
10Napkin: How Many ServersSize a service fleet from peak throughput, per-instance capacity, and explicit operating headroom.InteractiveConcept7 min20 XPOpen
11Find the Single Point of FailureRemove single serving failures by combining redundant front doors and services with a routable database failover path.DiagramConcept10 min20 XPOpen
12Blast RadiusReason about server, rack, zone, and region failures from where replicas share a failure boundary.InteractiveConcept7 min20 XPOpen
13Two of Everything?Prioritize redundancy at the remaining singleton whose loss would break the promised user path.InteractiveConcept7 min20 XPOpen
14Autoscaling, HonestlyAccount for detection and startup delay when pairing autoscaling with headroom or load shedding.InteractiveConcept7 min20 XPOpen
15Read the Traffic ChartDistinguish the first scaling-threshold crossing from the later peak that determines required capacity.InteractiveConcept8 min20 XPOpen
16The Far-Away UserExplain how regional routing reduces distance latency and why data placement still constrains the answer.InteractiveConcept7 min20 XPOpen
17Deploys Without DowntimeSequence traffic draining, health validation, and instance replacement for a rolling deploy without interruption.InteractiveConcept7 min20 XPOpen
18Scale the BakeryDesign a bakery service whose balanced instances share sessions and one durable order record.DiagramProject10 min20 XPOpen
19Explain HorizontalExplain how balancing healthy stateless instances raises capacity without tying a user to one process.InteractiveConcept8 min20 XPOpen
20Scaling Out Mastery CheckIntegrate redundant traffic distribution, interchangeable services, shared sessions, and replicated data in one scaled path.DiagramMastery check10 min40 XPOpen

The Request's Journey Caching