Year: 2024
Type: dissertation
Abstract: In the world of cloud computing, the serverless computing paradigm has rapidly gained popularity in recent years. Promising automatic scaling, usage-based billing and rapid development and deployment, it is an attractive cloud architecture approach for many use cases, such as spatial data processing. This thesis explores the possibility of using serverless computing to modernize the spatial data processing pipelines of Julkaisupalvelu, a spatial data publishing service. More specifically, the study aims to determine if serverless functions on the Microsoft Azure cloud platform can be used to replace the existing, virtual machine based data processing architecture. Both technical viability and cost-effectiveness are evaluated. To measure the feasibility of a serverless approach, a prototype serverless solution was designed, tested and analysed. The solution was implemented with Azure Functions, hosted on Azure Container Apps, and integrated with other Azure resources of Julkaisupalvelu. Testing with real-life spatial data was conducted to determine technical viability and usage data was gathered to estimate costs of the new solution. The results of the testing are promising. The serverless functions were able to outperform the current virtual machine solution by a significant factor, while providing a more scalable platform. At the same time, the new solution was also estimated to be cheaper. However, there were problems with function scaling and storage space that need to be solved before the solution can be adapted. At the end of this thesis, possible future research directions are provided to solve the issues in this study.
Source: Aaltodoc (Aalto University)
Author: Hyvönen, Tomas
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7034104915