Scholarlink

<p>Resource allocation and management in Fog-based Distributed Software-Defined Networks (Fog-DSDN) poses significant challenges due to their decentralized structure and the need for coordination across heterogeneous network layers. This study introduces two key innovations to address these challenges and optimize resource management. The first innovation is a novel neural model called GCSODE, which integrates a State-Space Ordinary Differential Equation (SS-ODE) framework with Graph Convolutional Networks (GCN) inside a Generative Adversarial Network (GAN) architecture. While SS-ODE effectively models continuous-time dynamics under irregular and non-uniform sampling, GCN captures structural dependencies among network nodes. This hybrid model leverages the GAN’s generative capability to mimic realistic data distributions and enables the dynamic adaptation of resource allocation based on both local and global observations collected by the SDN central controller. By processing the data received from distributed Fog nodes, the controller computes optimal allocation values and propagates them across the network. This model leads to significant improvements in latency reduction, energy efficiency, and response performance, thereby enhancing overall system efficiency. The second innovation involves the development of a globally federated learning mechanism based on tensor decomposition. In this approach, the SDN controller analyzes decentralized data from Fog nodes, extracts latent patterns, and generates globally optimal resource values that are distributed back to all nodes. This federated mechanism not only maintains data privacy but also ensures holistic coordination across the entire Fog-DSDN infrastructure. The proposed method, named GCSOT-DSDN, jointly captures spatial and temporal features through the integration of GCN and SS-ODE components. Extensive simulations using OMNET +  + under various network scenarios demonstrate that GCSOT-DSDN achieves superior performance, reducing response latency by 13.83%, improving load balancing by 24.74%, and decreasing energy consumption by 7.41%, compared to state-of-the-art baseline methods.</p>

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

A novel state-space ODE framework with tensor decomposition for resource allocation in fog-DSDN networks

  • 期刊论文

  • 发表日期：

    2026年2月12日

查看全文
  • Marjan Mahmoudi,
  • Behrang Barekatain,
  • Zahra Beheshti,
  • Alfonso Ariza Quintana,
  • Mohammad Rasoul Velayati

摘要

Resource allocation and management in Fog-based Distributed Software-Defined Networks (Fog-DSDN) poses significant challenges due to their decentralized structure and the need for coordination across heterogeneous network layers. This study introduces two key innovations to address these challenges and optimize resource management. The first innovation is a novel neural model called GCSODE, which integrates a State-Space Ordinary Differential Equation (SS-ODE) framework with Graph Convolutional Networks (GCN) inside a Generative Adversarial Network (GAN) architecture. While SS-ODE effectively models continuous-time dynamics under irregular and non-uniform sampling, GCN captures structural dependencies among network nodes. This hybrid model leverages the GAN’s generative capability to mimic realistic data distributions and enables the dynamic adaptation of resource allocation based on both local and global observations collected by the SDN central controller. By processing the data received from distributed Fog nodes, the controller computes optimal allocation values and propagates them across the network. This model leads to significant improvements in latency reduction, energy efficiency, and response performance, thereby enhancing overall system efficiency. The second innovation involves the development of a globally federated learning mechanism based on tensor decomposition. In this approach, the SDN controller analyzes decentralized data from Fog nodes, extracts latent patterns, and generates globally optimal resource values that are distributed back to all nodes. This federated mechanism not only maintains data privacy but also ensures holistic coordination across the entire Fog-DSDN infrastructure. The proposed method, named GCSOT-DSDN, jointly captures spatial and temporal features through the integration of GCN and SS-ODE components. Extensive simulations using OMNET +  + under various network scenarios demonstrate that GCSOT-DSDN achieves superior performance, reducing response latency by 13.83%, improving load balancing by 24.74%, and decreasing energy consumption by 7.41%, compared to state-of-the-art baseline methods.