📝 System Design Interview Classics · Deployment · Pro
Notification Service Design
A notification system that sends email, SMS and push messages at scale: one API, a queue per channel, user preferences, and retries.
Deployment diagrams are part of Pro. Anyone can view this one; generating and editing it needs Pro.
Drawing diagram…
What this diagram shows
- Separate queues so a slow SMS provider doesn't delay push
- User preferences and quiet hours are checked before sending
- Failed sends are retried with backoff, then parked in a dead-letter queue
Prompt used
Notification service deployment: internal services call the notification API on Kubernetes. The API checks user preferences in PostgreSQL, renders templates and puts messages on per-channel Kafka topics. Email, SMS and push workers consume their topics and call SendGrid, an SMS gateway and Firebase Cloud Messaging. Failures go to a retry topic and then a dead-letter topic.
Mermaid code
flowchart TB
SVC[Internal Services] --> API
subgraph K8S[Kubernetes Cluster]
API[Notification API]
TPL[Template Renderer]
subgraph WORKERS[Workers]
EW[Email Worker]
SW[SMS Worker]
PW[Push Worker]
end
end
subgraph KAFKA[Kafka]
TE[email topic]
TS[sms topic]
TP[push topic]
TR[retry topic]
DLQ[dead-letter topic]
end
PREF[(Preferences DB - PostgreSQL)]
API --> PREF
API --> TPL
API --> TE
API --> TS
API --> TP
TE --> EW
TS --> SW
TP --> PW
EW --> SG[SendGrid]
SW --> SMS[SMS Gateway]
PW --> FCM[Firebase Cloud Messaging]
EW -.failure.-> TR
SW -.failure.-> TR
PW -.failure.-> TR
TR -.after 5 tries.-> DLQRelated templates
URL Shortener System Design
The classic interview design for a service like bit.ly: short code generation, fast redirects from a cache, and click analytics processed separately.
Chat App Message Delivery (WhatsApp-style)
How a one-to-one chat message is delivered in a WhatsApp-style system, with sent, delivered and read ticks, and push notifications when the receiver is offline.
Social News Feed (Twitter-style)
How posts reach followers' feeds: fan-out on write for normal users, fan-out on read for celebrities, and a ranked feed built from a cache.
Ride Matching System (Uber-style)
The core of a ride-hailing app: drivers stream their locations, a geo index finds nearby drivers, and a matching service offers the trip to the best one.
API Rate Limiter (Token Bucket)
How a token bucket rate limiter decides whether to let an API request through, using Redis so all servers share the same counts.
Search Autocomplete (Typeahead)
How search suggestions appear as you type: a prefix index built offline from past searches, served from memory, and refreshed regularly.