Requirements

Functional Requirements:

• Create a short URL for a given long URL.
• Return the long URL associated with a given short URL.

Non-Functional Requirements:

• Low latency:
  • DB replication according to geo positioning
  • Cache, CDN for publicly exposed shortened url
  • Service replications in geo edge locations, managed by load-balancer (indicated as api gateway in diagram)
• High Availability
  • DB sharding, replication according to geo position
  • load-balancer for service replications
  • Service replicas in geo edge locations
  • a geo routing in the api gateway, to ensure resilience and high availability
• Scalability
  • handled by load balancer
  • scales up/down services in geo locations according to demand
• Reliability/resilience
  • handled by load balancer
  • reroutes to replicas at other geo locations if one goes down
  • for database, sharded replicas are there

API Design

• Post - for url creation
• Get - for fetching url mapping

High-Level Design

Flow goes like the following steps

• Client request goes to api gateway
• The api gateway decides which service, and which replica to send the request to, in a load balanced way
• For url creation
  • It generates a mapping
  • Sends the data to storage layer
  • Storage layer confirms saved data
  • api returns a response
• For url redirection
  • It asks storage layer for an existing mapping
  • If there is, it returns info
  • The redirect url is put in the header, and response is 302 (maybe), so the browser automatically redirects
• The storage layer
  • For creation, it saves the record in database directly
  • For read, it queries the cache first, if the data is there, then done
  • Otherwise gets from database, saves it in cache, then returns

Detailed Component Design

Cache

• Need analysis of record hits in cache. Just relying on new read may cause over-usage between cache-database transactions

Creation Service

• Multiple creation for same redirection may happen. As there is no user management, it is hard to track. So, database can be overflooded.

Latency

• Hard to have geo based storage, as a shortened url may be accessed from anywhere. Solution is, replication in multiple geo-edge locations, then analysis of hit locations to define where it should be