数据智能驱动的泛在Web应用服务质量优化方法研究

人机物融合的泛在计算环境对Web应用服务质量提出了新的要求和挑战。现有Web应用服务质量优化方法通常针对特定场景、采用启发式预设静态规则来配置和调度计算和网络资源以实现优化目标，适应性不强、灵活度不高，难以满足泛在Web应用的服务质量要求。.本项目以数据智能的视角，通过度量实验收集泛在环境下大规模Web应用运行数据，采用机器学习的技术手段构建Web应用服务质量预测模型，建立Web应用服务质量与应用静态结构和运行环境配置等特征之间的关系；进而可以根据此模型评估不同资源配置与调度策略下Web应用的服务质量，生成给定场景下的最优策略，最终通过部署该策略实现优化Web应用服务质量的目标。本项目将在Chromium浏览器内核上实现原型系统，并针对多个典型应用场景开展实验评估。

The ubiquitous computing environment of human-cyber-physical convergence puts forward new requirements and challenges to assuring the service quality of Web applications. Existing service quality optimization methods of Web applications usually use heuristic, predefined, and static rules to configure and schedule computing and network resources to achieve optimization goals for specific scenarios, which are not adaptable and flexible. As a result, it is difficult to meet the service quality requirements of ubiquitous Web applications..From the perspective of data intelligence, this project collects large-scale operating data of Web applications in ubiquitous environments through measurement experiments, uses machine learning technology to build a service quality prediction model, and establishes the correlation between Web application service quality, and application static structure and operating environment configuration According to this model, the quality of service of Web applications under different resource allocation and scheduling policies can be evaluated, and the optimal strategy in a given scenario can be generated. Finally, the goal of optimizing the quality service of Web applications can be achieved by deploying the generated policy. This project will implement a prototype system on the Chromium browser kernel and carry out experimental evaluations for multiple typical application scenarios.