Browsing by Author "Aarnio, Tuomas"

Now showing 1 - 2 of 2
  • No Thumbnail Available
    Item
    Applying event-driven architecture in a large-scale logistics system
    (2023-08-21) Aarnio, Tuomas; Rao, Balaji; Perustieteiden korkeakoulu; Haaranen, Lassi
    This thesis investigates the transition from traditional monolithic software architectures to event-driven models, emphasising their influence over system performance, scalability, and maintainability. A case study transitioning a large-scale logistics system to use event-driven architecture was conducted as a part of the thesis. The problem explication identified several problems inherent in the current system, including a shared database and the complexity of the existing codebase, largely due to technical debt accumulated over years of operation, which is common in long-lived projects. To address these challenges, an artefact was designed that applies an event-driven architectural approach. The artefact is responsible for managing the state of the deliveries, which is a central aspect of the logistics platform. Within the artefact, deliveries were modelled as state machines, with state updates managed through discrete events. This design choice enabled quick access and modification of the state data, improving the overall system performance. Apache Kafka was chosen as the event streaming platform. One of the significant challenges faced during this transformation was the integration of the new artefact into the existing system. A migration plan was formulated where initially only direct write queries related to the delivery state were replaced with events. A temporary service was created to ensure synchronization between the new artefact and the shared database, allowing the query of the delivery state from both the shared database and the event stream produced by the artefact. This approach facilitated a phased refactoring process, minimizing disruption to the existing services while enabling a gradual and controlled transition. Event-driven architecture turned out to be successful for this use case in terms of scalability and reduction of technical debt. The practical implementation posed challenges, and a good understanding of the underlying business domain for modelling the events and early and effective communication among stakeholders were crucial for the process.
  • No Thumbnail Available
    Item
    Flow-Based Generative Models
    (2021-04-16) Aarnio, Tuomas; Mesquita, Diego; Perustieteiden korkeakoulu; Hyvönen, Eero