Högskolan i Skövde

This study investigates the performance differences between REST and GraphQL, focusing on their efficiency in data fetching within different network environments, specifically in terms of WiFi and 4G mobile networks. As microservices and API architectures have evolved, REST has become a standard. However, GraphQL is a promising and interesting alternative, addressing REST issues such as over-fetching. 

This study aims to answer two primary research questions: 1) How does GraphQL perform in relation to REST with regard to response times, throughput, and payload size in both regular WiFi and 4G settings? And 2) in what scenarios is GraphQL most efficient? 

To answer these questions, a series of experiments was conducted using a locally developed GraphQL server and a REST API, both benchmarked under various user load scenarios using Apache JMeter (for the rest of the paper to be referred to as “JMeter”). 

The findings indicate that GraphQL demonstrates an ability to reduce payload sizes and decrease response times and throughput, particularly in mobile network environments. It is also shown that it matters how GraphQL is implemented in terms of what type of source the data is fetched from, demonstrating significant differences between services such as Azure Cosmos DB and Elasticsearch.