User-Centered Performance Engineering of Model Transformations
User-Centered Performance Engineering of Model Transformations
复制标题
DOI：
10.1109/models-c.2019.00097
复制
发表时间：
2019-09
期刊：
影响因子：
--
通讯作者：
Raffaela Groner
中科院分区：
文献类型：
--
作者：
Raffaela Groner
在模型驱动工程中，模型是关键的工件。由于待开发的系统变得越来越大、越来越复杂，相应的模型也变得越来越大、越来越复杂。这种趋势也会影响这些模型上的操作，比如转换。它们在设计时和运行时被应用，例如更新模型、生成代码或创建新模型。随着模型大小的增加，它们的执行时间也会增加，这使得它们的性能成为一个重要的质量方面。目前的研究主要集中在进一步改进执行转换的转换引擎，但这并不能单独解决问题。引擎优化永远不可能缓解所有可能的性能问题，因为有任意数量的方法来定义转换以及影响运行时的模型和元模型。因此，转换工程师还必须确保他们以一种执行时间短的方式定义转换。为了实现这一点，模型转换的性能工程方法是必要的。这种方法必须包含有助于分析和改进性能的步骤和技术。在本文中，我们提出了声明性模型转换的性能工程方法。我们确定了构成我们方法的五个工件：指导方针、监视、分析、可视化和改进建议。这些工件旨在帮助工程师在基于监视的分析和可视化的帮助下理解转换的执行和性能问题的原因，以便改进它们。在改进过程中，工程师将得到指导方针和改进建议的支持。
In Model-Driven Engineering, models are key artifacts. Due to the fact that the systems to be developed become larger and more complex, the corresponding models also become larger and more complex. This trend also influences operations on these models, such as transformations. They are applied at design time and at runtime, e.g. to update models, generate code or to create new models. With increasing model size, their execution time increases, making their performance an important quality aspect. Current research mainly concentrates on further improvements of the transformation engine that performs the transformation, but this will not solve the problem alone. Engine optimizations will never be able to mitigate every possible performance problem due to the fact that there's an arbitrary amount of ways to define a transformation as well as the models and meta-models that all affect the runtime. Therefore, transformation engineers must also ensure that they define their transformations in such a way that they have a short execution time. To achieve this, a performance engineering approach for model transformations is necessary. This approach must consist of steps and techniques that help to analyze and improve performance. In this paper we present our performance engineering approach for declarative model transformations. We identified the five artifacts Guidelines, Monitoring, Analyses, Visualizations and Improvement proposals that form our approach. These artifacts are intended to help an engineer to understand the execution of a transformation and the causes of performance problems with the help of Analyses and Visualizations based on our Monitoring in order to improve them. During the improvement the engineer will be supported by Guidelines and Improvement proposals.