Continuous and proactive software architecture evaluation: An IoT case -- Dataset generated from iFogSim

There will always be a trade-off between using the simulators and physical IoT devices in experimentation and data generation. This is due to the high cost of the actual deployment of IoT devices as compared to simulators. However, some companies, such as Amazon, IBM, and Intel, are motivating the need for having IoT simulation instrumenting what-if test scenarios, typically used during the architecture analysis and refinement stages to evaluate the response and sensitivity of the architecture to these tests. Additionally, many researchers are currently looking for an IoT dataset that provides QoS for IoT architectures. This work provides a dataset well-tested for the most important quality attributes when evaluating IoT architectures. In particular, this work used iFogSim to generate QoS of various IoT architectures in the form of Response Time, Energy consumption, and network usage. After that, MOA framework was used to generate the Forecast QoS values using different time series forecasting algorithms.在实验与数据生成过程中，使用模拟器与实体物联网（IoT, Internet of Things）设备始终存在权衡关系。这是因为相较于模拟器，实体物联网设备的实际部署成本高昂。不过，亚马逊（Amazon）、IBM、英特尔（Intel）等企业推动了物联网模拟的相关需求，即构建假设测试场景——此类场景通常应用于架构分析与优化阶段，用于评估架构针对此类测试的响应特性与敏感度。此外，当前诸多研究者都在寻求能够为物联网架构提供服务质量（QoS, Quality of Service）指标的物联网数据集。本项研究构建的数据集已针对评估物联网架构时的核心质量属性完成充分测试。具体而言，本工作借助iFogSim生成了各类物联网架构的服务质量指标，涵盖响应时间、能耗及网络使用率三类维度。随后，本研究通过MOA框架，结合多种时间序列预测算法生成了服务质量预测值。