· Created a Multithreaded Java Client to upload a day of item purchases from multiple stores in different time zones and exert various heavy concurrent loads on the server, utilized different strategies to increase the throughput
· Built the sever to accept requests and store the data, then Load balancing incoming network traffic across multiple servers
· Scaling out the purchase database to enable Read/Write Splitting, shard the purchases data among instances
· Achieved asynchronous & stream processing by using RabbitMQ & Kafka broker, to reduce latencies
· Created a Microservice which also receives and records purchase data in H2 database, served to answer queries about store
In the final version: The client used 1024 threads to simulate 1024 stores. For each store, it will send 5 items per purchase, 300 purchases per hour and 9 hours pur day. The server side used four ec2 instances and a four partitioned kafka broker The database is sharded into four rds instances The throughput exceeded 10000, which is 10 times better than the first server version
Final Result page
