Atelier product profile is a microservice that was scaled on AWS EC2 instance using 2 host servers, caching and load balancing.
Development: 
Database:
Cloud Side: 
For testing, the most complex select query in this microservice, styles, is chosen. Tests are benchmarked at 1000 clients per second.
Two databases considered were Postgresql and Mongodb. I chose Postgresql because the data I have is structured and highly relational. The biggest reason why I went with postgresql was because this microservice is read-only and with Postgres indexing, I could optimize my queries to be around 0.1 ms.
K6 was used as the testing tool. The goal was to have average response time to be under 50ms.
Microservice was deployed to an Amazon EC2 instance that was located in the US West coast. Before caching, and with only one host 1000 rps had a high latency at around ~3326ms and ~1.3% of the requests were timing out.
Adding a second server and balancing the load using the server with least connection method drastically changed the latency, bringing it down to ~460ms. Though the error rate was still high.
Changed the load balancer's configuration to keep 20 connections alive at a time, even though it brought down the error rate, it was still above the desired min, 1%.
Finally added caching using Nginx. As a result, queries response latency went down to ~62ms and error rate to 0%.
Possible next steps to increase the throughput would be to add more servers. Another step would be to upgrade to more powerful machines.
Needs npm install. Check out package.jsonfor further details.