Bachelors thesis on GraphQL API performance

This repo contains the practical part of my bachelors thesis, which is about optimizing performance of GraphQL APIs, targeting pagination algorithms and the N+1 problem.

The complete set of test results acquired during the making of the thesis can be found in the test-results.xlsx file.

1. Setup

1.1 Prerequisites

We'll need to install Node.js 16 and setup a PostgreSQL 13 database.

1.2 Clone the repo

Clone the repo and run npm i in /api, /scripts and /tests

1.3 Setup the database

Make sure that the PostgreSQL server is running and check the connection details in /api/ormconfig.js.

Then, in a terminal, move to /api and run npm run dev to start the backend for the first time.

This will connect to the database and set up all tables.

1.4 Acquiring and importing the data

Grab the electronics metadata file from http://deepyeti.ucsd.edu/jianmo/amazon/index.html (make sure to cite them if used scientifically). Then unzip the json file, rename it to data.json and move it to /scripts/data/in.

Then open a terminal in /scripts and run npm run data:prepare to convert it to csv files. You can find them in /scripts/data/out once the process is complete.

Lastly, import those csv files into your database with a tool such as DataGrip. Make sure to have your import tool strip out the header row, as well as escape double quotes.

2. Run the tests

Now that the API is usable, you can open a terminal in /tests and use npm start to run the tests. You don't want to start the API manually, the test process will do that for you.

After the tests are done, you'll find generated CSV files in /tests/src that contain all measurements for each test. Open them in Excel or another tool of your choice.

3. Starting the API manually to test against it

If you need to start one of the API implementations to run queries against it, you can do so by moving to /api and use npm run build to build it, and then one of the following commands to start any of the implementations:

• npm start -- --impl offset to start the offset-based implementation
• npm start -- --impl cursor-offset to start the cursor-based implementation with offset support
• npm start -- --impl cusror-offset-dataloader to start the dataloader backed version of the cursor-based implementation

You will then be able to test the API either with a tool such as postman/insomnia, or simply by accessing GraphiQL on http://localhost:3000/