System Design

Overview

The intention of this project is to build a horizontally scaled back end for an API serving a front end e-commerce store's microservice.

Requirements

• Service must respond to over 1000 requests per second (RPS) with a sub 2% error rate and sub 1s response time on a database containing 30 million generated records.

Final Architecture

Key Metrics Achieved

• 1300 RPS on AWS EC2 micro instances
• 0.1% error rate

Data Generation

Generated and wrote 30 million records to a CSV file on disk using a recursive algorithm.

Front End Deploy

• Statically hosted microservices across 4 AWS CloudFront CDN instances

Load Balancing

• Used Nginx as a load balancer utilizing a least stressed server approach

Server Hosting

• Deployed 6 Dockerized Node.js servers to AWS EC2 instances in order to reduce processing time
• Increased query performance and processing time through simplification of queries
• Increased client pool to 600 clients

PostgreSQL

Performance Improved using the following strategies

• Denormalization of data for faster query times
• Indexing of keys used for search
• Partitioned data into 1 million record segments to increase speed of lookup
• Increased maximum clients in pool to 600
• Increased working memory to maximum possible on AWS EC2 micro instances

Testing

• K6 for load testing locally and in the cloud
• New Relic for visual metrics across each node server and Post
• Loader.io for client reliability testing