Design and evolution of highly adaptive software systems: Context, control and run-time validation

随着web服务和社会技术生态系统的快速发展，这些现代的、分散的、分布式计算系统的复杂性给企业带来了巨大的挑战。从以商品为中心的业务到以服务为中心的业务的不断演变需要新的和创新的方法来构建、运行、管理和发展软件系统。终端用户越来越多地要求企业提供灵活、有弹性、基于位置、面向服务、分散、节能、自修复和自优化的智能软件系统。具有这种高度自适应特性的系统必须能够在运行时根据其对环境和自身状态的感知以全自动或半自动自适应的形式调整其行为。

With the rapid growth of web services and socio-technical ecosystems, the complexity of these modern, decentralized, distributed computing systems presents significant challenges for businesses. The continuous evolution from goods-centric to service-centric businesses requires new and innovative approaches for building, running, managing, and evolving software systems. End-users increasingly demand that businesses provide smart software systems that are flexible, resilient, location-based, service-oriented, decentralized, energy-efficient, self-healing, and self-optimizing. A system with such highly adaptive properties must be capable of adjusting its behaviour at run-time in response to its perception of its environment and its own state in the form of fully or semi-automatic self-adaptation. In this research we aim to investigate innovative control and context driven self-adaptation to produce smarter applications. One of the most promising approaches to achieving such properties is to equip software systems with feedback control to address the management of inherent system dynamics. Applications range from smart web services and location business intelligence to adaptive cloud scheduling and system diagnosis. Our innovative approach to orchestrate run-time system adaptations is based on dynamic context and control theory and involves run-time verification and validation. We will validate our results with real-world systems by collaborating extensively with industry. In this process, we will train many graduate students as well as industrial collaborators. This research is timely and of critical strategic importance for our nation. More autonomicity in computing systems will lead to simpler systems, better resiliency, more stability, higher availability, fewer errors, and higher security. Investigating how highly adaptive systems evolve over time and developing a framework for evaluating control architectures will not only reduce the difficulties with long-lived self-managed systems, but also provide guidance on how to equip existing applications with highly adaptive components.