Scholarlink

The Cloud-Edge Continuum represents a complex computational environment where data processing must be dynamically distributed between localized edge devices and centralized cloud resources. This paper presents the results of the MPESACC project, proposing an end-to-end framework enabling the automatic transformation of sequential Python code into distributed workflows. Our approach introduces a directive-based programming model utilizing Python decorators to define computational intent, constraints, and data dependencies without refactoring the underlying logic. This metadata drives a static analysis and scheduling engine that employs a Constrained Heterogeneous Earliest Finish Time (HEFT) algorithm to optimize resource allocation. Furthermore, the framework integrates CLOWNSim (Cloud Workloads Naive Simulator), a discrete-event tool designed to evaluate hierarchical deployment strategies and loop-based algorithmic expansion. Experimental results demonstrate that this adaptive scheduling approach can reduce execution time by up to 55% compared to monolithic edge execution, confirming the effectiveness of the proposed methodology in distributed environments.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Methodologies for the Parallelization, Performance Evaluation and Scheduling of Applications for the Cloud-Edge Continuum: From Design Principles to Experimental Validation

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Antonio Esposito,
  • Rocco Aversa,
  • Enrico Barbierato,
  • Maria Carla Calzarossa,
  • Beniamino Di Martino,
  • Luisa Massari,
  • Daniele Tessera,
  • Salvatore Venticinque,
  • Luca Zanussi

摘要

The Cloud-Edge Continuum represents a complex computational environment where data processing must be dynamically distributed between localized edge devices and centralized cloud resources. This paper presents the results of the MPESACC project, proposing an end-to-end framework enabling the automatic transformation of sequential Python code into distributed workflows. Our approach introduces a directive-based programming model utilizing Python decorators to define computational intent, constraints, and data dependencies without refactoring the underlying logic. This metadata drives a static analysis and scheduling engine that employs a Constrained Heterogeneous Earliest Finish Time (HEFT) algorithm to optimize resource allocation. Furthermore, the framework integrates CLOWNSim (Cloud Workloads Naive Simulator), a discrete-event tool designed to evaluate hierarchical deployment strategies and loop-based algorithmic expansion. Experimental results demonstrate that this adaptive scheduling approach can reduce execution time by up to 55% compared to monolithic edge execution, confirming the effectiveness of the proposed methodology in distributed environments.