Cloud-to-Edge Computing: A Unified Computational Paradigm

物联网(IoT)涉及大量生产和消费数据的互联设备。在过去的几年里，物联网在学术界和许多行业都获得了极大的欢迎。物联网设备可以在许多环境中找到，例如智能家居和智能建筑、智能车辆、医疗设备、工业4.0等。Gartner估计，到2020年，将有超过200亿台物联网设备。*历史上，物联网设备主要由简单的传感器和执行器组成，它们在云中转发数据，几乎不需要本地处理。如今，物联网设备变得越来越强大和互联，并正在迅速超越其低级嵌入式设备的传统角色，变得类似于嵌入式计算机。例如，广受欢迎的Raspberry Pi Zero W设备配备了1 GHz CPU、512MB内存、WiFi和蓝牙支持，仅需5美元。随着这些设备变得更强大，它们可以运行完整的操作系统(例如，Linux)和用高级、程序员友好语言(例如，Java、Java、Python)编写的应用程序，这为直接在这些设备(即，边缘计算)上运行更复杂和更高级的应用程序和服务打开了大门。*我提出的研究旨在提供一个全面的模型和统一的范例，允许开发人员以统一和可互操作的方式编写大规模分布式高级边缘应用程序，并将这些应用程序及其使用的服务(例如，数据存储、通信、文件系统)在靠近数据产生的位置动态部署，并直接部署到物联网设备本身。我相信，这可以产生几个好处，并有助于缓解纯粹集中式(即云)模式的限制，例如减少带宽和基础设施成本；减少延迟，这可能会影响实时数据处理和决策；以及减少对中心故障点的依赖。*拟议的研究有可能探索并提供新的解决方案，以应对前沿研究领域(云端计算)的多维挑战。研究成果将通过向顶级系统会议(如ICDC、SEC、中间件)发表出版物的形式获得，此外还有助于培训几(7)名研究生，他们将学习重要的研究和领导技能，使其在蓬勃发展的学术和行业市场中具有价值。考虑到物联网环境和电信行业中的几个关键参与者(例如贝尔、Telus、思科、英特尔)以及加拿大本身的优势，可以预见对加拿大行业的好处；例如，应用和服务的分布式部署可能对地理上分散的社区和远程站点有利，并可以确保在本地直接在受信任的设备上处理敏感数据。**

The Internet of Things (IoT) involves large amounts of interconnected devices that produce and consume data. Over the last few years, the IoT has gained tremendous popularity both in academia and many industry sectors. IoT devices can be found in many contexts, such as smart homes and smart buildings, intelligent vehicles, medical devices, the industry 4.0, etc. Gartner has estimated that there will be over 20 billion IoT devices by the year 2020.******Historically, IoT devices consisted mostly of simple sensors and actuators that relayed data to/from the cloud with little to no local processing. Nowadays, IoT devices are getting more and more powerful and connected, and are quickly evolving beyond their traditional role of low-level embedded devices to become similar to embedded computers. For instance, the popular Raspberry Pi Zero W device features a 1 GHz CPU, 512 MB of RAM, WiFi and Bluetooth support, for only 5 USD. As these devices become more powerful, they can run full operating systems (e.g., Linux) and applications written in high-level, programmer-friendly languages (e.g., Java, JavaScript, Python), which opens the door to running more complex and higher-level applications and services directly on these devices (i.e., edge computing).******My proposed research aims at providing a comprehensive model and a unified paradigm that will allow developers to write large-scale distributed high-level edge applications in a unified and interoperable manner, and to dynamically deploy these applications and the services that they consume (e.g., data storage, communications, file systems) close to where the data is produced, and directly onto the IoT devices themselves. I believe that this can yield several benefits, and can help alleviating the limitations of a purely centralized (i.e., cloud) model, such as reducing bandwidth and infrastructure costs; reducing latency, which can have an impact on real-time data processing and decision making; and alleviating the dependence on a central point of failure.******The proposed research has the potential to explore and provide novel solutions to multi-dimensional challenges in a bleeding-edge field of research (cloud-edge computing). Research outcomes are expected through the form of publications to top-tier systems conferences (e.g., ICDCS, SEC, Middleware), in addition to contributing to the training of several (7) graduate students, which will learn important research and leadership skills that will make them valuable in thriving academic and industry markets. Benefits to the Canadian industry are envisioned, given that several key players are well established in the IoT landscape and in the telecom industry (e.g., Bell, Telus, Cisco, Intel), as well as benefits to Canada itself; e.g., a distributed deployment of applications and services could be beneficial in geographically dispersed communities and remote sites, and could ensure that sensitive data is processed locally, directly on trusted devices.**