Using model-driven engineering to support autonomic computing

软件复杂性急剧增加，并且仍在继续增加。因此，开发和维护软件系统的成本也大大增加。自主计算（AC）旨在构建具有自配置、自管理、自保护、自修复和自优化（自 * 属性）能力的系统，以减少人为错误、系统停机时间并提高系统效率。然而，自主系统的发展是复杂的。模型驱动工程（MDE）在开发生命周期中系统地使用模型作为主要的软件工件。高级模型可以转换为更详细的模型，并可能最终生成源代码。MDE技术有可能降低构建和验证自主系统的成本。换句话说，可以使用各种模型（例如，软件模型、策略模型、服务质量（QoS）模型和性能模型），这些模型可以根据需求使用工具进行验证，并且可以在其他生命周期阶段使用。然而，通常很难自动化或修改行为模型。此外，大多数现有的AC MDE方法依赖于明确已知的状态，这对于某些应用可能是不切实际的。本研究的主要目标是：（1）扩展现有的生成方法以支持自适应特性，从而更好地理解自适应及其在性能和QoS方面的相关代价;（2）评估现有的自主计算体系结构模型，扩展自主计算体系结构分析方法;（3）制定决策战略，以支持在动态变化的环境中何时、如何适应;以及（4）开发模型符号、转换和管理技术，以促进AC的MDE。系统复杂性是软件开发和维护中的一个重要问题，AC被用来解决系统复杂性问题。拟议的项目提供了潜在的解决方案，以改善AC的MDE技术，这反过来又有助于降低自主系统的开发和维护成本。

Software complexity has increased dramatically and it still keeps increasing. As a result, the cost of developing and maintaining software systems has also increased significantly. Autonomic computing (AC) is aimed at building systems that have the ability to self-configure, self-manage, self-protect, self-heal, and self-optimize (self-* properties) to reduce human errors, system down time, and improve system efficiency. However, development of autonomic systems is complicated. Model-driven engineering (MDE) applies systematic use of models as primary software artifacts during the development life cycle. High-level models can be transformed into more detailed models and possibly the eventual generation of source code. MDE technologies have the potential to reduce the cost of building and validating autonomic systems. In other words, autonomic systems can be described using various models (e.g., software model, policy model, quality-of-service (QoS) model, and performance model) which can be validated with tools against the requirements and can be utilized at other lifecycle phases. However, it is often difficult to automate or revise the behavioral model. In addition, most existing MDE approaches for AC rely on explicitly known states, which may be impractical for some applications. The main objectives of the proposed research is to (1) extend our existing generative approach to support self-* properties to better understand self-adaptation and its associated cost in performance and QoS; (2) evaluate existing autonomic architectural models and extend architecture analysis methods for autonomic computing; (3) develop decision making strategies to support when, what, and how to adapt in dynamically changing environments; and (4) develop model notations, transformations, and management technologies to facilitate MDE for AC. AC has been advocated to address system complexity which has become a crucial issue in software development and maintenance. The proposed project offer potential solutions to improve MDE technologies for AC which in turn could help to reduce the development and maintenance cost of autonomic systems.