Edge View SNDlib Database formatted for VNE_CRS

5G technologies have enabled new applications on a heterogeneous and distributed infrastructure edge which unifies hardware, network and software aimed at digital enabling. Based on the requirements of Industry 4.0, this infrastructure is developed using the cloud and fog computing sharing model, which should meet the needs of service level agreements in a convenient and optimized way, requiring an orchestration mechanism for the dynamic resource allocation. Among these mechanisms, virtual networks embedding (VNE) and dynamic resource management (DRM) have shown a way to define where and how edge technology should be used. This paper proposes a resource allocation algorithm, VNE_CRS, which uses an artificial intelligence technique called reinforcement learning to orchestrate multiple domains, benefiting from its characteristic of considering the whole problem, end-to-end, using different aspects of 5G Quality of Service Indicator (5QIs). Experiments were carried out in simulation comparing VNE_CRS with state-of-the-art algorithms for the multi domains Edge environment. Results have shown that the usage of reinforcement learning techniques to VNE resource allocation has shown performance gains. It can not only simplify the VNE architecture but also act as a full orchestration system that aims to the strategic long run results of whole infrastructure usage.

第五代移动通信技术催生了在异构和分布式的基础设施边缘的新应用，该基础设施融合了硬件、网络和软件，旨在实现数字化赋能。基于工业4.0的需求，该基础设施采用云计算和雾计算共享模型进行开发，旨在以方便和优化的方式满足服务级别协议的需求，并需要一种动态资源分配的编排机制。在这些机制中，虚拟网络嵌入（VNE）和动态资源管理（DRM）为确定边缘技术应如何及在哪里使用提供了可行路径。本文提出了一种资源分配算法VNE_CRS，该算法利用强化学习这一人工智能技术来编排多个领域，得益于其全面考虑整体问题、端到端、利用5G服务质量指标（5QIs）不同方面的特性。在仿真实验中，将VNE_CRS与适用于多领域边缘环境的现有算法进行了比较。结果表明，将强化学习技术应用于VNE资源分配，在性能上取得了显著提升。这不仅简化了VNE架构，还作为一种全面的编排系统，旨在实现整个基础设施使用的战略长远效果。