Design a URL Shortener

You own a link shortener. It creates short codes and redirects them — that part works.

What does not work is the scale it is about to operate at. Ten billion redirects a month averages roughly 3,800 per second, and traffic is not evenly distributed: peaks run several times the average, and a small number of links account for most of the volume.

The read:write ratio is about 100:1. That asymmetry is the single most important fact about this system, and your architecture should show that you noticed it.

You start with an API tier querying PostgreSQL directly. Every redirect is a database read.

Requirements

URLs created: 100M / month
Redirects served: 10B / month (~3.8K/s average)
Peak traffic: ~40K/s
Read:write ratio: 100 : 1
p99 redirect latency: < 100 ms
Availability: > 99.99%

Service targets

p99 < 100ms86.7ms

Availability > 99.99%0.06%

Database CPU < 70%99%

Scored on trade-offs, including how your design behaves at traffic you have not tried yet.

Components

Edge

Compute

Data

Async

Click to add. Drag from a node’s right edge to connect it.

3 components2 connections0:00

drag to move · delete to remove · double-click to rename

Traffic40Kreq/s

Timeline

Recording…

Traffic

40Kreq/s

p50

45ms

p99

86.7ms

p99 < 100ms

Errors

0.06%

Availability > 99.99%

Dropped

24req/s

Cost

$534/mo