Harokopio University
Towards a self-healing IoT event-based architecture: infusing self-optimized real-time data imputation and quality management to promote fault tolerance
Abstract
dc:descriptionThis thesis addresses a central reliability problem in event-driven IoT systems: pipelines may remain operational while silently losing evidential integrity as sensor events become missing or delayed. Such data-quality degradation propagates through windowed aggregation and complex event processing (CEP), undermining derived event products and the decisions that depend on them. The thesis argues that dependable event-driven IoT requires accountable continuity: sustaining real-time outputs under imperfect evidence while making the degree and consequences of degradation explicit rather than concealed. Grounded in a Design Science Research methodology, the thesis develops and evaluates an integrated self-healing approach spanning architectural design, quality-aware event transformation, and real-time repair. First, it proposes a self-healing, event-based architecture that treats event transformation as a first-class, composable concern and provides monitoring and orchestration hooks for quality-driven adaptation. Second, it introduces quality-aware CEP mechanisms that tolerate missing and late events through controlled inclusion and event fabrication, coupled with explicit characterization of completeness and timeliness so that downstream consumers can interpret outputs in context. Third, it establishes a production-like experimental methodology based on controlled fault injection and repeatable telemetry capture to quantify availability–quality trade-offs under realistic degradation regimes. Finally, it advances edge-side real-time imputation under strict cadence and resource constraints through an interpretable, self-optimized model selection strategy that adapts repair decisions to observed conditions, improving tolerance-compliant continuity beyond any single static method. Two applied case studies, including smart greenhouse monitoring and a weather-station edge pipeline, demonstrate that the proposed approach increases the availability of derived event products under missingness while preserving transparency about quality degradation and maintaining feasibility for constrained deployments. Overall, the thesis contributes a coherent set of architectural principles, mechanisms, and empirical guidance for building event-driven IoT systems that remain operationally useful under imperfect evidence, without sacrificing interpretability or accountability.
Degree
thesis:*- Grantor dc:publisher
- Harokopio University
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Gkoulis, Dimitrios
- Γκουλής, Δημήτριος
Subjects
dc:subject × 34- Συμβανοκεντρικό Διαδικτύο των Πραγμάτων
- Αυτοϊάσιμα συστήματα
- Αρχιτεκτονική μικροϋπηρεσιών
- Επεξεργασία σύνθετων συμβάντων με επίγνωση ποιότητας
- Μεταδεδομένα ποιότητας δεδομένων
- Πληρότητα και χρονική εγκαιρότητα
- Ελλείποντα και καθυστερημένα συμβάντα
- Κατασκευή συμβάντων
- Υπολογιστική αιχμής
- Συμπλήρωση ελλειπόντων τιμών σε ροές δεδομένων σε πραγματικό χρόνο
- Ερμηνεύσιμη επιλογή μοντέλου κατά τον χρόνο εκτέλεσης
- Συμβιβασμός διαθεσιμότητας–ποιότητας
- Διαδίκτυο των πραγμάτων
- Event-driven iot
- Self-healing systems
- Microservices architecture
- Quality-aware complex event processing
- Data-quality metadata
- Completeness and timeliness
- Missing and delayed events
- Event fabrication
- Edge computing
- Real-time stream imputation
- Interpretable runtime model selection
- Availability–quality trade-off
- Internet of things
- Επιστήμη Ηλεκτρονικών Υπολογιστών και Πληροφορική
- Φυσικές Επιστήμες
- Κυβερνητική
- Επιστήμη ηλεκτρονικών υπολογιστών, θεωρία και μέθοδοι
- Computer and Information Sciences
- Natural Sciences
- Cybernetics
- Computer science, theory and methods
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/61162