115,000 Read QPS / 1,150 Write QPS
Sustained peak request volume during high-traffic events.
High-ThroughputMid Level
High-Scale URL Shortener & Key Generation
Design a globally distributed TinyURL service with 100:1 read-heavy traffic and sub-10ms redirects.
Target Scale
Engineering Scale & Performance SLAs
Target production parameters expected in a senior or staff interview round.
100 Million Active Links
Daily active users generating read and write operations.
15 TB storage / 5 years
Projected data ingestion and replication storage capacity.
Redirect < 10ms
Strict end-to-end percentile latency SLA constraint.
Stage 01
Functional & Non-Functional Requirements
Establish clear problem boundaries before proposing architectural components.
Core System Capabilities
- Generate unique 7-character short URLs from arbitrary long destination URLs.
- Redirect short URLs to original URLs with HTTP 301/302 redirects.
- Support custom alias short links and expiration TTLs.
Reliability & Latency SLAs
- High availability (99.999% uptime) with zero downtime deployments.
- Ultra-low redirect latency (p99 < 10ms).
- Collision-free unique key generation.
Stage 02
Capacity Estimation Math
Step-by-step arithmetic conversions for QPS, storage, and bandwidth.
| Dimension | Calculation Formula | Estimated Result |
|---|---|---|
| Write & Read Throughput | 100M new URLs/month = ~40 write QPS. 100:1 read ratio = ~4,000 read QPS (Peak = ~115k QPS) | ~1,150 Peak Write QPS / ~115,000 Peak Read QPS |
| 5-Year Storage Capacity | 100M URLs/month × 12 months × 5 years × 500 bytes/record | ~30 Billion URLs = ~15 Terabytes total persistence |
| Cache Memory Size | 20% hot URLs account for 80% read traffic. 115k QPS × 86,400s × 0.2 × 500B | ~100 GB RAM Redis Cache Cluster |
Stage 03
Multi-Tier Architecture & Component Topology
How requests navigate ingress gateways, application logic, caching, and persistence.
Dedicated Key Gen Service · Zookeeper / Etcd Coordinator
Pre-generates Base62 7-character keys into memory buffers, avoiding on-the-fly hash collisions.
Nginx / Envoy Load Balancer · Go / Rust Stateless Redirect Service
Handles short code lookups from cache, returning immediate HTTP 301 redirects.
Redis LRU Cache · Cassandra / DynamoDB Distributed Store
Stores key-to-URL mappings sharded by the first 2 characters of the short key.
Stage 04
Database Schemas & Partitioning Strategy
Entity models, indexing, and primary key partitioning.
PK: short_key
- short_key (VARCHAR, PK)
- original_url (VARCHAR)
- user_id (VARCHAR)
- created_at (TIMESTAMP)
- expires_at (TIMESTAMP)
Partitioned by short_key hash with secondary index on user_id for analytics.
Stage 05
Critical Architectural Trade-Offs
How to defend engineering compromises when challenged by interviewers.
HTTP 301 vs HTTP 302 Redirect
Option A: 301 Permanent Redirect (Browser caches, reduces server load)
Option B: 302 Temporary Redirect (Server intercepts every click, tracks analytics)
Rationale: Selected 302 for commercial analytics and link revocation, with edge CDN caching to balance latency.
MD5/SHA256 Hashing vs Key Generation Service (KGS)
Option A: Hash + Base62 Encoding (Requires collision resolution loops)
Option B: Stand-alone Key Generation Service (Pre-generates unique sequential keys)
Rationale: KGS eliminates duplicate hash retries and guarantees O(1) instantaneous write latency.
Technical FAQ
Frequently Asked Questions: High-Scale URL Shortener & Key Generation
Key interview questions and conceptual defenses.
Why use Base62 encoding for short URLs?
Base62 uses [a-z, A-Z, 0-9]. A 7-character string provides 62^7 = 3.5 trillion unique URLs, easily sufficient for decades of scale.
Simulate this architecture
Download ClawPad