System requirements

Functional:

1. URL shortening service
2. Redirect using short URL
3. Custom alias for a long URL (optional)
4. Expiration of URL

Non-Functional:

1. Highly available

Capacity estimation

More reads than writes 10:1 (reads: writes)

1 million active users for a day

12000 Reads per second

1200 writes per second

API design

1. Post /Shortner
   1. Requests a URL along with userId, expiration date(optional), and custom alias(optional)
   2. Responds with a short URL, Expiration date, and custom alias

Database design

1. High reads and low writes, and a key value pair, so a NoSQL database and it is high avalilable.
2. userId, short URL, original URL, expiration date, custom alias

High-level design

User UI -> LoadBalancer -> Application servers (url shortner generator) connected to a database.

For reading, use Redis for caching the most-recently URLs. We can also add analytics service

Request flows

user inputs a long URL, URL generator service will generates a short URL and stores in the database and redirects it using the redirecting service.

Detailed component design

URL generator sevice: takes the original long URL. Converts it into a md5 hash whichis 32bit hex.

With the generated hex a short URL is generated and also checks for duplicates. Then checks stores it in the database. If user provided a custom alias, then this service will check if that alias is already in the database or not. If it is in the database returns a error that the URL is already in use. else it stores and maps it to the original URL

Trade offs/Tech choices

For it to be highly available we create shards for the database and horizontally scaled the databases. So Here we follow eventual consistency.

Failure scenarios/bottlenecks

Must be able to solve the hash collisons which might occur if two long urls produce same md5 hash.

Expired URLs must be removed from the database and should not support any redirects.

Future improvements

What are some future improvements you would make? How would you mitigate the failure scenario(s) you described above?

1. In orderto handling collisons we can use -1, -2 at the end of that hash
2. We can have a background task or cron job to regularly check the DB na ddelete any expired URLs