Department of Computer Science: MSc Thesis Presentations

Analysing learning management system performance

Author: Jimmy Ihalainen
Supervisor: Professor Petri Vuorimaa

Abstract: Modern internet-scale applications are often broken up into small independently operated services and distributed across millions of computers around the globe in datacenters. In these microservices architectures, the small independent services communicatethrough a lightweight mechanism, such as an HTTP API. Due to the nature of these types of architectures, traditional observability tools, such as logs and simple metrics are no longer sufficient at tracking transactions and explaining the full picture of thesystem. The number of processes involved in every end-user request is simply too large. Distributed tracing is the answer to many of the difficulties caused by the massive increase in complexity that these modern distributed architectures entail.

This thesis implements distributed tracing into the A+ learning management system using open-source frameworks, such as OpenTelemetry and Grafana Tempo. The platform has previously suffered from performance issues due to the lack of performance testing tools.This thesis investigates how distributed tracing can be utilized to improve the observability of A+ with the goal of giving more insight into its performance. Performance testing tools were developed to help with analysing the service performance in differentscenarios before rolling out updates to the production environment. The results describe the benefits and limitations of distributed tracing when attempting to analyse the performance of a microservice-based platform such as the A+.