Integration of system and software architecture techniques in support of autonomic cloud management

对球场的看法。需求工程（RE）和系统/软件架构（SSA）中的技术对于达到一定的质量水平以及支持有效的管理和高效的发展是至关重要的。分区在RE和SSA中起着核心作用。文献中提出了各种RE和SSA划分技术。尽管如此，这些划分技术主要是基于引出的需求生成静态系统结构。已经提出了一些方法来考虑响应需求、技术或环境变化的动态软件体系结构。然而，随着云计算（CC）、虚拟化方法（如虚拟网络、虚拟机和虚拟存储）和软件定义范式（如自定义网络、自定义数据中心和自定义环境）等新兴技术的出现，SSA面临着新的挑战，因为这些新兴技术可能会动态改变SSA。例如，由于资源管理、服务水平协议（sla）服务质量（QoS）的原因，系统的某些部分可能会动态迁移并运行在由云服务提供商（CSP）甚至多个CSP管理的不同云（多云或公共云与私有云）上，这可能会违反一些用户约束或策略，例如隐私、安全、信任（PST）或成本。（用户可以是企业或组织。）现有的RE和SSA模型被认为不适合CC，特别是软件即服务（SaaS）。因此，传统的RE和SSA技术必须适应不断发展的技术。* *的研究目标。该研究的总体目标是研究新兴的CC范式，并适应现有的RE和SSA相关技术，以应对SaaS中的挑战。具体来说，我打算研究与SaaS和适合SIMT的建模技术密切相关的新兴软件定义范例。要做到这一点，首先需要系统地研究CC的传统RE和SSA技术，这些RE和SSA技术包括划分、特征建模、模型驱动工程（MDE）、面向方面设计（AOD）、自主计算、软件性能工程（SPE）、资源管理和分布式计算。我一直在RE和SSA领域工作。然而，这些技术必须加以扩展和/或调整，以适应CC的新兴技术和挑战。*本建议的主要短期目标如下：请注意，这些目标是相互关联的，它们不必以严格的顺序方式进行。*(1)对SaaS的传统RE和SAA建模方法进行表征，并对这些方法进行扩展以适应SaaS需求；*(2)为SIMT开发一种集成软件定义技术和增强的RE和SAA建模技术的方法；*(3)配置SaaS的运行时行为，并开发有效的跨层（SaaS、PaaS和IaaS）云管理和约束满足（例如QoS和PST的约束）的方法。**此外，主要的长期目标包括：(4)研究模型符号并开发模型转换方法，以支持SaaS领域的MDE；(5)开发一种机制，便于将用户约束映射到云系统约束。这两个长期目标都需要坚定的基础研究和正式方法的调查。从长远来看，拟议的研究将有助于知识库和HQP机会，以支持高效和有效的SaaS开发，并简化不同CC技术（包括IaaS、PaaS和SaaS）的集成。

Perspective on the field. Techniques in requirements engineering (RE) and system/software architecture (SSA) are crucial to achieve a certain level of quality and to support effective management and efficient evolution. Partitioning plays a central role for RE and SSA. Various RE and SSA partitioning techniques have been proposed in the literature. Nonetheless, those partitioning techniques mostly generate a static system structure based on elicited requirements. A few approaches have been proposed to consider dynamic software architecture in response to changes in requirements, techniques, or environments. However, as the advent of emerging technologies, such as Cloud Computing (CC), virtualization approaches (e.g., virtual networks, virtual machines, and virtual storage), and software-defined paradigms (e.g., self-defined networking, self-defined data centres, and self-defined environments), there are new challenges in SSA, as SSA may be dynamically changed as a result of those emerging technologies. For instance, parts of a system may be dynamically migrating and running on different clouds (multiple clouds or public vs. private clouds) that are managed by the cloud service provider (CSP) or even multiple CSPs for resource management, service-level agreements (SLAs) quality-of-service (QoS) reason, which may violate some user constraints or policies, e.g., privacy, security, trust (PST), or cost. (Users could be enterprises or organizations.) Existing RE and SSA models are identified inadequate for CC, especially Software-as-Service (SaaS). Therefore, traditional RE and SSA techniques have to be adapted to accommodate the evolving technologies.**Research Objectives. The overall goal of the research is to investigate emerging CC paradigms, and adapt existing RE and SSA related technologies to address the challenges in SaaS. Specifically, I intend to investigate emerging software-defined paradigms that are closely related to SaaS and the modeling techniques that are suitable for SIMT. To do that, there is a need to first systematically investigate traditional RE and SSA techniques for CC. Those RE and SSA techniques include partitioning, feature modeling, model-driven engineering (MDE), aspect-oriented design (AOD), autonomic computing, software performance engineering (SPE), resource management, and distributed computing. I have been working on those RE and SSA areas. Those technologies, however, have to be expanded and/or adapted to meet emerging technologies and challenges of CC.* The main short-term objectives of this proposal are listed as follows. Note that those objectives are inter-related and they do not have to be carried out in a strict sequential manner. *(1) Characterize the traditional RE and SAA modeling methods for SaaS and expand those methods to accommodate SaaS demands; *(2) Develop an approach that integrates software-defined technologies and the enhanced RE and SAA modeling technologies for SIMT; and*(3) Profile run-time behaviors for SaaS and develop an approach for efficient cross-layer (SaaS, PaaS, and IaaS) cloud management and constraint satisfaction (e.g., constraints for QoS and PST).**In addition, the primary longer term objectives include: (4) Investigate model notations and develop model transformation methods to support MDE in the area of SaaS; and (5) Develop a mechanism that facilitates mapping of user constraints to cloud system constraints. Both longer term objectives require firm fundamental research and investigation in formal methods. In the longer term, the proposed research will contribute to the knowledge base and HQP opportunities that support efficient and effective SaaS development and ease integration of different CC technologies, including IaaS, PaaS, and SaaS.