Requirements

Functional Requirements:

List the key functional requirements for the system (Ask the AI for hints if stuck)...

• Should accept a long form link
• Should output a short format link
• Should map the correct input links to output links through a relational database

Non-Functional Requirements:

• Links should be validated with a ping of sorts to ensure invalid links don't seep through
• Application should scale up/down on basis of user traffic
• Application should be highly available
• API response times should be fast
• Highly viral links should be cached so tons of requests don't hit the server

Capacity Estimation

Daily active users - 10,000 users

Daily links - 100k+

Uptime - 98%

API response time - 100ms

API Design

POST /create-link - should input a long-format link and output a short format link

GET /:link - should redirect users to the actual link

High-Level Design

• Use load balancer and Kubernetes to scale up/down on the basis of load
• Use SQL to ensure unique constraints for ShortenedLink column and take advantage of indexes for fast retrieval.
• Use Redis cache on top for O(1) retrieval.
• Create 2 services for LinkGenerator (POST /link) and LinkRetriever (GET /:linkSuffix)
• LinkGenerator PINGS the actual link before to check for 200 status; then stores it in DB as well as cache.
• LinkRetriever hits the cache, and then DB to get the actual link and redirect from there. If cacheKey doesn't exist, create one.
• Can utilize read replicas for high availability of DB.
• Use a edge CDN to cache viral links so server isn't flooded with same requests.
• Need a CreateShortLink service that uses a combination of timeStamps, workerId, and GUIDs to ensure unique shortLink even with distributed DBs and missed caches.
• Multi-level caching and cache-miss management - CDN stores response -> misses or TTL expired -> server cache -> misses or TTL expired -> hit DB and create cache

Database Design

Use a relational DB to ensure unique constraints and take advantage of indexes for fast retrieval

Links table schema:

ShortenedLink | ActualLink | CreatedBy

Detailed Component Design

• LinkGenerator - the POST API component that takes an input and shortens the URL before validation checks.
• CacheSettings - Use a long-lived TTL of 24 hrs for CDN and Redis to ensure fast responses over a long period of time.
• ID Generator - the component that uses a combination of timestamps, workerIds, and GUIDs to generate a guaranteed unique ID for each created link. Having unique worker IDs for each cluster ensures DB stays consistent across split-brain and generator-outage scenarios.