System Design : Ad Click Aggregator

https://www.youtube.com/watch?v=6TroztUV3f8&list=PLjTveVh7FakLGZ36GbWAk_DMf_0xBZpGv&index=4

https://docs.google.com/presentation/d/1sx1oxYRYNwvhEoPM7LqgXPWaZI5kv64-139hNdLpwB4/edit?slide=id.g3707c69fac3_0_10#slide=id.g3707c69fac3_0_10

Requirement : Functional

• Every Click of each advertisement should be counted
• We should be able to run arbitrary analytical queries on the data

Requirement : Functional

• Secure
• Data Privacy Compliance
• Call Volumes
• High Availability
• Low Latency : query should be returned as fast as possible
• time series granularty of 1 min
• FAST : Data should be available withing a few seconds of click

High level

• since APIs are stateless , round robin load balancer can be used
• check duplicate for malicous users
• batch data as parquet files and push in data warehouse

its not a good choice to use dynamo DB instead of kafka , may not be able to support write throughput

Kafka has log based writes , so it always appends data , hence disk moves sequentially not randomly

Kafka one partition can easily do 50k writes per second per partition

Data / Entity Object diag ( he did not do it ) 

• linkClicked(String ad-unique-id, String user-id, timestamp)  

Block Diagram