Scholarlink

<p>The growing diversity and volume of IoT services make efficient and QoS-aware service selection a critical challenge, particularly as existing discovery mechanisms mainly rely on functional parameters and iterative decision-making processes that scale poorly under large service registries and high request loads. While several approaches accelerate selection through pre-filtering or clustering, they often sacrifice relevance and remain unable to handle user-defined QoS priorities with varying weights. To address these limitations, this paper introduces MLQoSI, a Multi-Level Quality of Service Index designed to structure IoT service repositories according to QoS profiles explicitly expressed in user requests. MLQoSI groups similar requests into classes by profile similarity and maintains, for each class, a dynamically updated list of promoted services ranked with TOPSIS. A monitoring threshold controls when recalculations are required, ensuring an effective balance between stability and responsiveness. The approach is fully integrated into an IoT architecture and evaluated through extensive simulations on datasets comprising 12,000 services and 40,000 user requests. Results show that MLQoSI significantly reduces service retrieval time–achieving up to 30x&#xa0;speedup compared to iterative TOPSIS–while maintaining high and stable relevance above 90%, outperforming methods based on Skyline, PCA-clustering, and other MCDM strategies. These findings demonstrate that MLQoSI provides a scalable, accurate, and QoS-aware service selection mechanism suitable for large-scale IoT environments.</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 QoS based multilevel index for an efficient service selection in the IoT

  • 期刊论文

  • 发表日期：

    2026年6月2日

查看全文
  • Meriem Achir,
  • Abdelkrim Abdelli

摘要

The growing diversity and volume of IoT services make efficient and QoS-aware service selection a critical challenge, particularly as existing discovery mechanisms mainly rely on functional parameters and iterative decision-making processes that scale poorly under large service registries and high request loads. While several approaches accelerate selection through pre-filtering or clustering, they often sacrifice relevance and remain unable to handle user-defined QoS priorities with varying weights. To address these limitations, this paper introduces MLQoSI, a Multi-Level Quality of Service Index designed to structure IoT service repositories according to QoS profiles explicitly expressed in user requests. MLQoSI groups similar requests into classes by profile similarity and maintains, for each class, a dynamically updated list of promoted services ranked with TOPSIS. A monitoring threshold controls when recalculations are required, ensuring an effective balance between stability and responsiveness. The approach is fully integrated into an IoT architecture and evaluated through extensive simulations on datasets comprising 12,000 services and 40,000 user requests. Results show that MLQoSI significantly reduces service retrieval time–achieving up to 30x speedup compared to iterative TOPSIS–while maintaining high and stable relevance above 90%, outperforming methods based on Skyline, PCA-clustering, and other MCDM strategies. These findings demonstrate that MLQoSI provides a scalable, accurate, and QoS-aware service selection mechanism suitable for large-scale IoT environments.