1 of 13

Enhancing Verification in the �IoT-Edge-Cloud Continuum

Davide Taibi

University of Oulu

2 of 13

Context and Motivation

  • Distributed IoT-edge-cloud systems are dynamic and heterogeneous

​

  • Architectural drift and erosion are common in complex systems

​

  • Traditional test-based verification is insufficient

​

  • We need verification that accounts for architecture and runtime behavior

3 of 13

Software Architecture in the Continuum

4 of 13

Architectural Reconstruction

  • Goal: Reconstruct actual system architecture from artifacts for comparison against design-time models

​

  • Static: call graphs, dependency analysis
  • Dynamic: runtime traces, observability data

​

5 of 13

Architectural Compliance Testing

  • Identify conformance violations (e.g., layering breaches, illegal calls)
    • Use reconstructed architecture as test oracle
    • Automate rule checking (e.g., service isolation, gateway boundaries)

​

  • Goal: Compliance as code using model comparisons

6 of 13

Verification Strategy Across Layers

  • Combine static (structure), dynamic (behavior)
    • Edge: test under constraints, local decision logic
    • Cloud: verify orchestration, data flows
    • IoT: ensure adherence to contracts and communication interfaces

7 of 13

Generative AI for Verification

  • AI generates tests for uncovered execution paths
  • Suggests architecture rules from historical models
    • Limitations: false positives, runtime costs

​

  • Goal: LLM-assisted compliance suggestion and validation

8 of 13

6G and Software Architecture in the Edge-to-Cloud Continuum

  • Architectural Disruption by 6G
    • Shift to massively distributed and heterogeneous nodes
    • Requires architectural redesign to support ultra-low latency and >100 Gbps throughput
    • Convergence of IoT, AI, Quantum Computing, and edge orchestration

​

  • New Architectural Needs
    • Energy-Aware Orchestration (EAO) models with dynamic resource scaling
    • Resilient deployment models for smart hospitals, AR, connected vehicles
    • Integration of AI-driven diagnostics and liquid AI at the edge

​

  • Compliance & Sustainability
    • Necessity to rethink modularity, service isolation, and deployment granularity
    • Emphasis on sustainability, reliability, and AI-augmented architectural compliance
    • Enhanced architectural traceability and automation with LLMs and static/dynamic analysis

​

  • Goal: Build future-proof, adaptive architectures for real-time, high-stakes applications in 6G environments

9 of 13

Case Example: Smart City System

  • Components: sensors, edge gateways, cloud analytics
    • Issue: unauthorized data flow from edge to 3rd-party service
  • Approach: reconstruct actual call graph, detect drift
  • Outcome: compliance violation detected and patched

10 of 13

Case Example: FinTech Proactive Autoscaling

  • Components: Prometheus, Grafana, K3s (lightweight Kubernetes), Edge Cluster VMs
    • Production traces from a FinTech payment system
    • Observed bottlenecks during scale-up delays under peak CPU loads

​

  • Issue:
    • Default Kubernetes autoscaler reacts after CPU spikes;
    • Risk of under-provisioning and late pod instantiation
  • Approach:
    • Applied 29 time series forecasting models (14 statistical + 15 neural) to predict CPU usage;
    • Integration with k3s for proactive pod scheduling

​

  • Outcome:
    • All models ran inference in < 1 minute, -> compatible with real-time auto-scaling
    • Auto-scaler behavior refined, reducing premature scale-down and ensuring resource readiness

11 of 13

Current Status

  • Static analysis tools
    • Architectural Patterns
    • Anti-Patterns
    • Architectural Degradation (drifts)

​

  • Dynamic Analysis tools
    • Ongoing

​

  • Resource Specification and allocation
    • K8 model
    • Orchestration models / allcoations

​

​

12 of 13

Challenges & Opportunities

  • Challenges:
    • Trace-model alignment
    • Edge observability
  • Opportunities:
    • Self-updating architecture models
    • AI-assisted verification pipelines
    • Continuous architectural compliance testing

13 of 13

Conclusion & Vision

  • Verification must evolve to include architectural integrity

​

  • Combine static/dynamic analysis with AI and reconstruction

​

  • Enable adaptive, resilient verification in the IoT-cloud continuum

​

  • Let’s build automated compliance pipelines for the future