Explorar performance com Apollo Federation

• Main Author: Queirós, Leandro Daniel Oliveira
• Publication Date: 2023
• Format: Master thesis
• Language: eng
• Source: Repositórios Científicos de Acesso Aberto de Portugal (RCAAP)
• Download full: http://hdl.handle.net/10400.22/23985
• topic: GraphQL; Apollo Federation; Performance; GraphQL Federation

The growing tendency in cloud-hosted computing and availability supported a shift in soft ware architecture to better take advantage of such technological advancements. As Mono lithic Architecture started evolving and maturing, businesses grew their dependency on soft ware solutions which motivated the shift into Microservice Architecture. The same shift is comparable with the evolution of Monolithic GraphQL solutions which, through its growth and evolution, also required a way forward in solving some of its bot tleneck issues. One of the alternatives, already chosen and proven by some enterprises, is GraphQL Federation. Due to its nobility, there is still a lack of knowledge and testing on the performance of GraphQL Federation architecture and what techniques such as caching strategies, batching and execution strategies impact it. This thesis aims to answer this lack of knowledge by first contextualizing the different as pects of GraphQL and GraphQL Federation and investigating the available and documented enterprise scenarios to extract best practices and to better understand how to prepare such performance evaluation. Next, multiple alternatives underwent the Analytic Hierarchy Process to choose the best way to develop a scenario to enable the performance analysis in a standard and structured way. Following this, the alternative base solutions were analysed and compared to deter mine the best fit for the current thesis. Functional and non-functional requirements were collected along with the rest of the design exercise to enhance the solution to be tested for performance. Finally, after the required development and implementation work was documented, the so lution was tested following the Goal Question Metric methodology and utilizing tools such as JMeter, Prometheus and Grafana to collect and visualize the performance data. It was possible to conclude that indeed different caching, batching and execution strategies have an impact on the GraphQL Federation solution. These impacts do shift between positive (improvements in performance) and negative (performance hindered by strategy) for the different tested strategies.