Model-Driven Development for IoT Application Deployment

Abstract

The explosive growth of Internet of Things (IoT) ecosystems has exposed the fragility of hand-crafted deployment pipelines that rely on ad-hoc scripts, heterogeneous device configurations, and scattered YAML artifacts. Model-Driven Development (MDD) promises a principled alternative by raising the level of abstraction: architects capture structure, behavior, resources, and quality objectives as models, and transformations generate deployable artifacts for edge, fog, and cloud tiers. This manuscript proposes IoTDeployML, a lightweight domain-specific modeling approach with three coordinated viewpoints—Domain, Deployment, and Policy—and a transformation toolchain that produces microcontroller firmware stubs (C/MicroPython), Node-RED dataflow graphs, container images, and Kubernetes/Helm manifests. We formalize deployment as a constrained mapping problem between application tasks and a resource graph annotated with latency/energy budgets, then operationalize it using model-to-model (M2M) transformations and model-to-text (M2T) code generation.

A simulation study of a smart-campus scenario with 100 devices compares the proposed MDD pipeline against a baseline script-centric process. Across 30 randomized trials, MDD reduced time-to-deploy by ~60%, configuration errors by ~78%, service-level objective (SLO) violations by ~67%, and network bandwidth by ~24%; energy per message at the edge dropped by ~22%. Statistical analyses (normality checks, independent t-tests/ANOVA, effect sizes) show the improvements are significant (p < .01) with large effects. The paper details metamodels, transformation rules, orchestration hooks, and discusses threats to validity and practical adoption guidance.