README.MD

Test benchmark connection pooling in Go with Postgres

Author: kienmatu

Introduction

This is a simple project to test benchmark connection pooling in Go with Postgres. The purpose of this project is to compare the performance of three different methods of handling database connections:

• Using connection pooling
• Creating new connections each time
• And using a single connection.

Installation and Usage:

To run this project, you will need to have Go and docker installed on your system. Once you have installed these dependencies, you can follow the steps below:

Clone the repository:

git clone https://github.com/kienmatu/go-connection-pooling-test.git
cd go-connection-pooling-benchmark

Install the required dependencies:

go mod download

Start the Postgres server using Docker Compose:

docker-compose up -d

Wait for 1 minute for the data to be seeded.

Run the Sample API:

go run main.go

Install Go-wrk

go install github.com/tsliwowicz/go-wrk@latest

Test Benchmark

Now you can run the benchmark tests separately using go-wrk:

go-wrk -c 10 http://localhost:8080/products/pooled
go-wrk -c 10 http://localhost:8080/products/normal
go-wrk -c 10 http://localhost:8080/products/new

The results of each test will be displayed in the terminal.

Test elapsed time

Or you can test it with the index.html file:

How many connections is the best fit for pool size?

Actually, I set 50 and 90 for the first time bench, after that I realize that too much number of connections is not efficient.

The efficient number of connection pool size should be the same as your server/machine's physical number of cores.

I tested with 4 for both idle connections and max pool size, and the result nearly the same with 50 idle and 90 max conns.

I just also added the code in expressjs (but yeah, of course, the performance can not be strong as Go).