In this post, we explore the limits of single-node Postgres and what optimizations can extend its usefulness as a primary application database before

Exploring the limits of Postgres: when does it break?

submited by

Style Pass

2024-09-24 20:30:05

In this post, we explore the limits of single-node Postgres and what optimizations can extend its usefulness as a primary application database before requiring more advanced scaling solutions.

Let's use a real-world scenario, a monolith web application that uses PostgreSQL as the application database running on AWS RDS. The RDS instance is a 2xlarge.

In our ecommerce app, we display products for sale, where shoppers can add them to their carts and place orders. We want to support large amounts of concurrent users who are shopping, and avoid high latency for their requests that might reduce our sales.

Imagine we're doing architectural planning, and want to estimate what a 10x increase in concurrent users would look like with our current PostgreSQL instance. Let’s factor in the cost of server instances with a low and high value, so we can place our current usage in between.

We’ll use instances.vantage.sh to explore instances, gathering resource and cost information. We'll choose on-demand pricing, per month, in the US East region. The 2XL instance costs $650/month.