Intent-Based Network Management of Software Defined Wireless Sensor Networks

本研究的愿景是通过利用软件定义的网络框架来研究无线传感器网络中基于意图的策略。研究将集中在两个互补的研究领域：开发软件定义无线传感器网络（sdwsn）中分布式管理和控制的算法，以及通过开发语义策略来减少sdwsn网络管理中的策略差距。目标是减少移动传感和驱动场景（如身体传感器、车辆网络和网络物理系统的监控）中网络的能耗和延迟。这些场景是“新”物联网（IoT）服务的示例。也许物联网服务对传感器网络最具革命性的影响是改变这些传感器网络收集数据的方式，从简单的基于接收器的收集拓扑切换到基于流的拓扑。物联网服务要求来自分布式和连接传感器的更多数据能够近乎实时地提供。所有这些服务都在争夺同一资源。这方面的一个例子是来自车辆的数据，可用于更新车辆维护，交通替代路线和自动驾驶车辆的远程警告。所有这些服务都需要相同的数据，但具有不同的优先级和QoS。使用SDWSN可以更简单地管理这一点，但代价是网络需要更多的控制数据包开销。为了降低这种开销，本研究将研究无线传感器网络中的分布式SDN控制器，并结合传感器节点的集群形成。为了在传感器网络中保持QoS服务需求，需要开发一种基于意图的动态策略语言。目前，网络是使用特定于设备的策略来配置的，目标是使用“高级”的人类可理解的策略来配置这些策略，例如“要求在获取的1分钟内，每5分钟提供401号公路上车辆的平均速度。”在如何指定此需求并将其编程到网络中存在很大的差距，这被称为策略差距挑战。意图需要语义知识，这些知识可以弥合需求中使用的术语和用于配置网络的术语之间的差距。可以设想，现有的语义表示语言（如OWL和SWRL）可以用来创建高级意图策略，并弥补这种策略差距。关于物联网网络物理系统的全球市场，预计到2018年底，年收入将接近1360亿美元，加拿大市场将达到210亿美元。几乎所有主要的电信和IT公司都在改变他们的商业模式和网络，以适应这个面向服务的市场。预计机器对机器通信将对健康和制造领域产生重大影响，估计这两个领域约有75%的物联网服务。

The vision of this research is to investigate intent-based policies in wireless sensor networks by leveraging a software defined network framework. The research will focus on 2 domains of research that are complimentary: to develop algorithms to distribute management and control in Software Defined Wireless Sensor Networks (SDWSNs) and reduce the policy gap in network management of SDWSNs by developing semantic policies. The objectives are to reduce both energy consumption and delays in the network in mobile sensing and actuation scenarios such as the monitoring of body sensors, vehicular networks, and cyber-physical systems. These scenarios are examples of the "new" Internet of Things (IoT) services. Perhaps the most revolutionary impact that IoT services will have to sensor networks is to change the way in which these sensor networks collect data switching from a simple sink-based collection topology to a flow-based topology. IoT services are demanding more data from distributed and connected sensors to be available in near real-time. All these services compete for the same resource. An example of this is data from vehicles that can be used to update vehicle maintenance, traffic alternative routes, and long range warning for autonomous vehicles. All these services are requiring the same data but with different priorities and QoS. This can be managed simpler with an SDWSN at a cost of more control packet overhead to the network. In order to keep this overhead down this research will investigate distributed SDN controllers in the wireless sensor network in conjunction to cluster formations of the sensor nodes. In order to maintain the QoS service demands in the sensor network a language for intent-based dynamic policies will be developed. Currently networks are configured using device specific policies and the goal is to configure these using "high-level" human understandable policies such as "requiring the average velocity of vehicles on the 401 near the Oshawa every 5 minutes delivered within 1 minute of acquisition." There is significant gap in how this requirement is specified and programmed into the network and is known as the policy gap challenge. Intents require semantic knowledge that can bridge the gap between terminology used in the requirement and that used to configure the network. It is envisioned that existing semantic representation languages such as OWL and SWRL can be leveraged to create high-level intent policies and bridge this policy gap. Concerning the global market for IoT cyber-physical systems, it is expected to be nearly $136 billion in annual revenues and a $21 billion market in Canada by the end of 2018. Nearly all major telecom and IT companies are shifting their business models and networks to accommodate this service oriented market. Machine to machine communications is expected to have significant impact in the health and manufacturing domains with an estimate of about 75% of IoT services in those 2 domains.