Due to the explosive growth of smart devices, 5G, and the Internet of Things (IoT) applications in recent years, the volume and velocity of generated data, and consequently, delay-sensitive applications are increasing endlessly. This paper aims to improve the service delay and Quality of Service (QoS) by introducing HPCDF (Hybrid PSO-CRO Delay-improved for FogPlan) - an offline QoS-aware framework to deploy and release fog services dynamically. The proposed method provisions, i.e., deploy and release fog services to reduce service delay, based on the aggregated incoming traffic to each fog node. We formulate a cost function as an Integer Non-Linear Programming (INLP) problem by considering each service attributes, including required resources and associated traffic. This problem integrates storage, processing, deployment, communication costs, delay violation, high fog utilization reward, high traffic nodes cost, and service delay penalty. A hybrid binary PSO-CRO (Particle Swarm and Chemical Reaction Optimization) algorithm is proposed to achieve the lowest service delay and QoS loss to address this problem. The evaluation is performed on real-world traffic traces, provided by MAWI Working Group, under three different experiments to study the impact of various parameters of the hybrid binary PSO-CRO algorithm and the proposed framework on service delay. The evaluation results reveal that our proposed algorithm reduces service delay by 29.34%, service cost by 66.02%, and violates the delay 50.15% less in comparison to FogPlan framework.

翻译：近年来，随着智能设备、5G和物联网（IoT）应用的爆炸式增长，所生成数据的体量与速度，进而使时延敏感应用持续激增。本文通过提出HPCDF（针对FogPlan的混合粒子群-化学反应优化时延改进方法）——一种离线QoS感知框架，用于动态部署和释放雾服务，旨在改善服务时延与服务质量（QoS）。该方法基于每个雾节点聚合的入站流量来进行服务供应，即部署与释放雾服务以降低服务时延。我们通过考虑每个服务属性（包括所需资源和关联流量），将代价函数建模为整数非线性规划（INLP）问题。该问题整合了存储、处理、部署、通信代价、时延违规、高雾利用率奖励、高流量节点代价及服务时延惩罚。为解决此问题，提出一种混合二进制粒子群-化学反应优化（PSO-CRO）算法，以获取最低服务时延和QoS损失。基于MAWI工作组提供的真实流量轨迹，在三种不同实验场景下评估了混合二进制PSO-CRO算法及所提框架各参数对服务时延的影响。评估结果表明，与FogPlan框架相比，所提算法使服务时延降低29.34%，服务成本减少66.02%，且时延违规次数减少50.15%。