About

Hi there,
I would like to present a small challenge task. The main purpose is to create a URL shortener service. The service takes long URLs as input and provides shortened, unique aliases that redirect to the original URLs.

Start

The simplest way to run the service is by starting it in a Docker container:

docker-compose up

Call API

To use this service You can:

• Open the web console http://localhost:3000/api/v1/shortener
  Use the following mutation to add a new record:

    mutation {
       shortenUrl(originalUrl: "https://example17.com")
    }
  

• Use a VSCode plugin: Refer to the file http.http
• Use the curl utility:

  curl -X POST http://localhost:3000/api/v1/shortener \
  -H "Content-Type: application/json" \
  -d ' {"query": "mutation {\n  shortenUrl(originalUrl: \"https://example.com\")\n}\n"}'
  

• ...

System Design

Technology stack

Technology Stack The technology stack was limited and defined in the challenge task description.
For a production-grade version, I would suggest using AWS Lambda and DynamoDB.

Using GraphQL may seem overengineered for this use case.

Implementation

To make this system scalable, I chose to implement stateless services.
These services are designed as a single module for simplicity, ensuring ease of development and maintenance.

Simplification

For simplification, this service does not allow editing or deleting stored links.
If you try to shorten an existing URL, the service will return the already stored shortened link.

Choosing a Short URL Alias Algorithm

• Hashing
• Base62 Encoding

To avoid collisions and eliminate the need for additional database checks (which would lead to extra DB access) or advanced techniques like a Bloom Filter (which is definitely outside the scope of this task), I chose Encoding.

Distributed unique ID

To generate unique IDs that can be used across multiple nodes while preserving chronological order, a composite 64-bit ID is utilized.

ID Structure:

• Timestamp (41 bits): Represents the number of milliseconds elapsed since the beginning of 2025.
• Data NODE ID (4 bits): Allows for 2⁴ = 16 nodes (data centers).
• Machine ID (3 bits): Allows for 2³ = 8 machines per node.
• Sequence Number (8 bits): Enables each machine to generate up to 2⁸ = 256 unique IDs per millisecond.

This design ensures efficient and scalable ID generation, making it suitable even for high-load production environments.

Possible usecases

This containerized solution can be deployed on Kubernetes or other auto-scaling platforms in combination with a load balancer.

Since customers will primarily read rather than write data, it is recommended to integrate a caching system to improve performance.

Refeerence

In conclusion, I would like to recommend the book that inspired me:
System Design Interview by Alex Xu.