Search

Lessons (32)

• Capacity, Availability, Latency and CAP

  … a design chooses latency targets, availability targets and a partition-time consistency policy](/img/system-design-l6/capacity …

• Binary Search

  … If capacity C works, every bigger capacity works too, so the condition flips once. Binary search the capacity …

• Storage Engines & Database Operations

  … Bound pool size, reserve capacity for critical traffic, set deadlines and investigate long transactions or slow queries. Backups …

• Design a File Storage Service

  … part and final checksums Sharing Metadata-owned ACL/capability Device sync Versioned change feed/conflict policy Object-store …

• Design an E-Commerce Checkout

  … Handle unknown provider outcomes through reconciliation; expire reservations; isolate checkout capacity from bulk exports and marketing traffic.

• Architecture Patterns at Scale

  … Client-server Most products starting out Central service capacity Microservices Independent domain ownership/deployments Distributed operations and contracts …

• Real-Time, Pub/Sub, Webhooks and CDC

  … an event stream can feed pub-sub consumers, webhook delivery, server-sent updates and database change capture](/img …

• Arrays & Strings

  … mermaid flowchart LR A["capacity 4<br/ ■■■■ full"] -- push → copy to bigger B["capacity 8<br/ ■■■■■□□□"] B -- 3 …

• SQL vs NoSQL Databases

  … The CAP theorem : when the network between servers breaks (a P artition), a distributed database must choose between …

• Database Internals, Replication and Scaling

  … Replicas increase read capacity but add lag; route read-your-own-write flows carefully. Connection pools bound database …

• Views, Triggers & Stored Procedures

  … Transaction-control and return-value capabilities vary by engine, so name the product in an interview.

• Debug Latency Like an Investigator

  … Symptom Test first --- --- p50 and p99 rise Broad capacity, network, deploy, or dependency issue. Only p99 rises Queueing …

• Premiere-Night Video Streaming

  … every segment does not call auth. - Define degradation: cap bitrate, disable previews, or queue sessions. Heartbeat progress to …

• The L6 System-Design Thinking Loop

  … is never charged twice. Peak Average traffic hides capacity limits. 2M checkouts in five minutes. Failure window Finds …

• Pods

  … node that has this much free A hard cap at runtime Exceed CPU – The container is throttled (slowed …

• Indexes, EXPLAIN & Query Plans

  … reading most of a small table can be cheaper than index random I/O. Operational checklist Capture slow …

• Deploying & Operating Microservices

  … Simple, no extra capacity needed; old and new versions run side by side for a while. mermaid flowchart …

• Finding Service Boundaries with DDD

  … Practical heuristics for cutting services 1. Follow business capabilities , not technical layers. ✅ "Payments service". ❌ "Database service", "Validation service …

• Dynamic Programming

  … subset sum, 0/1 knapsack dp[capacity] or dp[i][capacity] Two strings LCS, edit distance, regex matching …

• I — Interface Segregation Principle

  … small, role-based interfaces ts // ✅ One interface per capability interface Printer { print(doc: Document): void; } interface Scanner { scan …

• L — Liskov Substitution Principle

  … every caller of Bird.fly() } Fix: model the capability separately. js class Bird { eat() {} } class FlyingBird extends Bird …

• Code Smells & Refactoring

  … If not, write a few characterisation tests that capture today's behaviour. 2. Take tiny steps: rename, extract …

• Monolith vs Microservices

  … small, independently deployable services , each owning one business capability and its own data , talking over the network. mermaid …

• Deployments, ReplicaSets & Rollouts

  … passes), then removes an old one, and repeats. Capacity never drops. ⚠️ Without a readiness probe , Kubernetes sends traffic …

• Kafka vs RabbitMQ & the Kafka Ecosystem

  … orders,public.outbox", "topic.prefix": "shop" } } Change Data Capture (CDC) with Debezium reads the database's transaction log …

• Topics, Partitions & Offsets

  … same key → same partition → guaranteed order. - Partition count caps consumer parallelism; you can add partitions but it remaps …

• Consumers & Consumer Groups

  … independently, each with its own committed offsets . - Partitions cap parallelism; extra consumers sit idle. - Commit after processing for …

• Serverless Architecture

  … 0 when nobody uses it Traffic spikes Plan capacity ahead ✅ Scales automatically Ops work Patching, scaling, monitoring hosts …

• Helm, kubectl Cheat Sheet & Best Practices

  … true capabilities: { drop: ["ALL"] } Plus: scan images, pull only from trusted registries, use NetworkPolicies, keep Secrets in a …

• Data in Microservices: Sagas & the Outbox Pattern

  … twice, so consumers must be idempotent . Change Data Capture (CDC) tools like Debezium read the database's transaction …

• Behavioral Patterns II: Chain, Template, Iterator, Mediator, Memento, Visitor

  … Before the boss fight you save; if you lose, you load the snapshot. Problem: capture an object's …

• Domain-Driven Design: Strategic Design

  … Charge card"] -- X1["🩷 Stripe"] -- E3["🟧 Payment captured"] E3 -- E4["🟧 Parcel shipped"] classDef ordering fill: dbeafe …