shortener.md
· 1.7 KiB · Markdown
Raw
# System Design Interview: Designing a URL Shortener
The URL shortener (think bit.ly) is a classic warm-up design question. Here's a structure that fits in 35 to 45 minutes.
## Requirements
- Functional: create a short link for a long URL; redirect short to long; optional custom alias and expiry.
- Non-functional: very read-heavy (often 100 reads per write), low-latency redirects, links must not collide.
- Rough scale: 100M new links a month means about 40 writes a second and 4,000 reads a second.
## Generating the short key
1. **Counter + Base62:** a globally unique ID (database sequence, or a Snowflake-style ID) encoded in Base62. Seven characters give 62^7, about 3.5 trillion keys. Predictable, so add a shuffle if guessing links is a concern.
2. **Hash and truncate:** MD5 or SHA-256 of the URL, then take 7 characters. Needs collision handling (retry with a salt).
3. **Pre-generated key service:** a separate service hands out unused keys in batches; there's no collision check on the hot path.
## Storage and API
- `POST /links {long_url}` returns `{short}`; `GET /{short}` returns a 301 or 302 redirect.
- 301 is cached by browsers (less load, but you lose click analytics); 302 keeps every hit visible.
- A key-value store (DynamoDB, Cassandra) keyed by the short code is a natural fit.
## Scaling reads
Put a cache (Redis) in front of the database; a small fraction of links gets most of the traffic. Add a CDN or edge redirects for the hottest links.
## Extras to volunteer
Analytics via an async event stream (Kafka), abuse and malware scanning on creation, and a TTL-based cleanup job for expired links.
---
*Disclosure: I'm on the team behind [Linkjob AI](https://www.linkjob.ai/), an interview assistant.*
System Design Interview: Designing a URL Shortener
The URL shortener (think bit.ly) is a classic warm-up design question. Here's a structure that fits in 35 to 45 minutes.
Requirements
- Functional: create a short link for a long URL; redirect short to long; optional custom alias and expiry.
- Non-functional: very read-heavy (often 100 reads per write), low-latency redirects, links must not collide.
- Rough scale: 100M new links a month means about 40 writes a second and 4,000 reads a second.
Generating the short key
- Counter + Base62: a globally unique ID (database sequence, or a Snowflake-style ID) encoded in Base62. Seven characters give 62^7, about 3.5 trillion keys. Predictable, so add a shuffle if guessing links is a concern.
- Hash and truncate: MD5 or SHA-256 of the URL, then take 7 characters. Needs collision handling (retry with a salt).
- Pre-generated key service: a separate service hands out unused keys in batches; there's no collision check on the hot path.
Storage and API
POST /links {long_url}returns{short};GET /{short}returns a 301 or 302 redirect.- 301 is cached by browsers (less load, but you lose click analytics); 302 keeps every hit visible.
- A key-value store (DynamoDB, Cassandra) keyed by the short code is a natural fit.
Scaling reads
Put a cache (Redis) in front of the database; a small fraction of links gets most of the traffic. Add a CDN or edge redirects for the hottest links.
Extras to volunteer
Analytics via an async event stream (Kafka), abuse and malware scanning on creation, and a TTL-based cleanup job for expired links.
Disclosure: I'm on the team behind Linkjob AI, an interview assistant.