Design and Management of Fog Networks

随着物联网(IoT)的出现，任何设备，无论多么小，都可以连接到互联网并监控物理元素，从而使智能城市、智能医疗和智能交通等应用成为可能。通常，这些设备生成的数据通过网络基础设施传输到云中进行处理/存储。然而，鉴于来自物联网设备的非凡数据的速度、分布和规模，处理云中的所有数据将是不切实际的。尤其是对于具有严格延迟要求的物联网应用，云计算无法充分满足这些要求。为了克服这些不足，出现了一种新的范式，称为雾计算。通过将计算(和存储)能力转移到用户附近，从而在物联网-雾-云连续体的任何位置处理数据，该范例已被作为一种有前景的解决方案引入以支持广泛的物联网应用；其目标是满足物联网应用的强制要求(例如，延迟)，并减少网络流量的爆炸性增长。近年来，雾基础设施在资源管理方面做出了显著贡献。事实上，已经提出了几种模型和方案来管理资源，目的是在最小化总体成本(例如带宽和能量)的同时满足物联网应用的要求(例如延迟)。然而，所有这些贡献都假定雾基础设施的存在。在这个项目中，我们将开发新的模型和工具，以高效地设计/规划物联网-雾-云连续体中的雾基础设施。更具体地说，我们将开发一个严格的数学公式，来研究和分析雾设计的动力学和特性。特别是，我们计划开发混合方法，将极其高效的元启发式方法与数学规划方法相结合，在减少组合爆炸的同时利用问题的结构。然后，我们将就关键设计问题进行研究：设计相对于负荷增加和随时间的地域扩展的可扩展性；以及配置对于故障的稳健性。然而，即使在最佳设计的情况下，超载和不平衡利用也会发生(例如，当实际负荷模式改变和/或超过设计阶段使用的估计/预期负荷时)。因此，我们还将开发新的自适应资源管理方案，以满足动态/不确定环境中的应用需求。更具体地说，我们将制定有效的方案，允许适应/重新优化FOG基础设施中当前支持的应用程序的资源供应，以应对这些不可预测的情况；目标是为这些应用程序提供服务质量保证。

With the emergence of Internet of Things (IoT), where any device, however small, is able to connect to the Internet and monitor/control physical elements, several applications were made possible such as smart cities, smart healthcare and smart transportation. Generally, data generated by these devices is transmitted, via the network infrastructure, to the cloud, to be processed/stored. However, with the rate, distribution and scale of extraordinary data from IoT devices, processing all the data in the cloud would be impractical. Particularly, when it comes to IoT applications with stringent delay requirements, cloud computing cannot adequately address these requirements. To overcome these deficiencies, a new paradigm, called fog computing, has emerged. This paradigm has been introduced as a promising solution to support wide range of IoT applications by moving the computing (and storage) power to the proximity of users, and thus processing data anywhere along the IoT-Fog-Cloud continuum; the objective is to meet the imposed requirements (e.g., delay), if any, of IoT applications and reduce the explosion in network traffic. Recently, there has been notable contributions in resource management of the fog infrastructure. Indeed, several models and schemes have been proposed to manage resources, with the objective to satisfy the requirements of IoT applications (e.g. delay) while minimizing the overall cost (e.g., bandwidth and energy). However, all these contributions assume the existence of the fog infrastructure. In this project, we will develop new models and tools for the efficient design/planning of fog infrastructure in the IoT-fog-cloud continuum. More specifically, we will develop a mathematically rigorous formulation, to study and analyze the dynamics and properties of fog design. In particular, we plan to develop hybrid approaches that combine extremely efficient meta-heuristics, with mathematical programming approaches that exploit the structure of the problem while curtailing combinatorial explosion. Then, we will carry out studies dealing with key design issues: the scalability of designs with respect to load increases and geographical expansion over time; and the robustness of configurations with respect to failures. However, even with an optimal design, overload and unbalanced utilization can and will happen (e.g., when actual load patterns change and/or exceed the estimated/expected load used during the design phase). Thus, we will also develop novel adaptive resource management schemes to satisfy application requirements in a dynamic/uncertain environment. More specifically, we will develop efficient schemes that allows to adapt/re-optimize resource provisioning in the fog infrastructure, of currently supported applications, in response to these unpredictable situations; the objective is to provide quality of service assurance to these applications.