System Design: Notification/Push Service

Requirements

• Multi-channel: mobile push (APNs/FCM), email, SMS; templates, preferences, retries, dedupe.

Architecture

Producers → Topic (Kafka) → Router → Channel Workers (APNs/FCM/SMTP/SMS) → Providers ↘ Store (Postgres) for receipts, logs, preferences

SLOs

• Enqueue→provider ACK P95 < 300 ms; Exactly-once user experience via dedupe keys.

Capacity

• 100k notifications/sec fanout bursts; workers horizontally scaled; provider quotas respected.

Failure modes

• Provider outage → circuit break and requeue with exponential backoff; channel fallback.

What Interviewers Look For

Notification Systems Skills

1. Multi-Channel Support
   • APNs/FCM for mobile push
   • Email/SMS channels
   • Template management
   • Red Flags: Single channel, no templates, poor management
2. Reliability & Delivery
   • Exactly-once semantics
   • Retry mechanisms
   • Deduplication
   • Red Flags: Duplicate notifications, no retry, unreliable delivery
3. Provider Integration
   • Provider quotas
   • Circuit breakers
   • Fallback strategies
   • Red Flags: No quota management, no circuit breakers, provider failures

Distributed Systems Skills

1. Message Queue Design
   • Kafka for high throughput
   • Topic partitioning
   • Consumer groups
   • Red Flags: Wrong queue choice, no partitioning, bottlenecks
2. Scalability Design
   • Horizontal scaling
   • Worker pools
   • Load distribution
   • Red Flags: Vertical scaling, bottlenecks, poor distribution
3. SLO Management
   • Latency targets (< 300ms)
   • Throughput targets (100k/sec)
   • Red Flags: No SLOs, high latency, low throughput

Problem-Solving Approach

1. Failure Handling
   • Provider outages
   • Network failures
   • Rate limiting
   • Red Flags: Ignoring failures, no handling, poor recovery
2. Edge Cases
   • Duplicate notifications
   • Provider quotas
   • Channel failures
   • Red Flags: Ignoring edge cases, no handling
3. Trade-off Analysis
   • Latency vs reliability
   • Cost vs features
   • Red Flags: No trade-offs, dogmatic choices

System Design Skills

1. Component Design
   • Router service
   • Channel workers
   • Preference store
   • Red Flags: Monolithic, unclear boundaries
2. Storage Design
   • Receipt tracking
   • Preference management
   • Logging
   • Red Flags: No tracking, no preferences, no logs
3. Monitoring
   • Delivery rates
   • Provider health
   • Latency metrics
   • Red Flags: No monitoring, no metrics, no visibility

Communication Skills

1. Architecture Explanation
   • Can explain notification flow
   • Understands multi-channel
   • Red Flags: No understanding, vague explanations
2. Reliability Explanation
   • Can explain retry logic
   • Understands deduplication
   • Red Flags: No understanding, vague

Meta-Specific Focus

1. Notification Systems Expertise
   • Multi-channel knowledge
   • Reliability focus
   • Key: Show notification systems expertise
2. Scale & Performance
   • High throughput design
   • Low latency
   • Key: Demonstrate scale expertise