Managing Data for the Internet of Things: Storage, Updates, and Transactions

考虑以下2016年6月《经济学人》杂志上的真实用例。汤姆·罗杰斯是加州中央谷的一位杏仁种植者，他必须不断监测许多变量，如湿度、营养成分、健康威胁和温度，并采取适当措施控制这些变量的值，为此他必须付出高昂的代价。汤姆想通过提高种植缺水杏仁所需的水和肥料的总体利用率来降低运营成本。Tom还希望访问有关他的农场状态的实时信息，他可以使用这些信息来做出决策，以实现预期的成本降低。Tom的问题是使用所谓的智能农业应用程序的一个完美示例，它是越来越多的应用程序跨越农业、交通（资产管理）、住房（智能建筑）和基础设施（智能电网）等不同领域的一个实例。这些应用的共同点是越来越多地使用物联网（IoT）来监控操作，以实现降低总体运营成本和提供有关被监控实体状态的实时数据的目标。物联网是一种网络范例，其中联网点（称为“事物”）是配备嵌入式连接的唯一可识别设备。为了实现他的既定目标，Tom在他的整个坚果种植园安装了无线传感器来测量上述变量。然后，测量结果定期发送到专用的第三方云服务器，计算出应该施用于种植园的适当的水和肥料水平。最后，这些结果被用来自动驱动农场的灌溉系统，向坚果林输送精确数量的水和肥料。归档在云数据库上的数据也由分析应用程序处理。我们提出的研究重点是设计和实现存储、更新和事务模型，用于管理物联网数据，这些数据从传感器流向云网关、云数据库、应用程序，然后返回。在短期内，我们将研究物联网数据管理的架构选择，并在云网关级别研究数据管理（例如，流清洗和融合，以及复杂事件检测），考虑到巨大的量，速度和种类的物联网数据需求。从长远来看，我们将研究云物联网分布式数据库的存储模型架构，以及适合物联网数据管理的更新和事务模型，并开发原型系统。我们预计，这项将由一家财富500强公司以及几位知名数据库和机器学习专家合作的研究，将对数据库行业的新方向、物联网应用以及物联网技术的改进具有重要意义。

Consider the following real-world use case presented in The Economist in June of 2016. Tom Rogers, an almond farmer in California Central Valley, must pay high costs by constantly monitoring many variables such as moisture, nutrient content, health threats, and temperature, and by taking appropriate actions to control the levels of the values of these variables. Tom wants to reduce operating costs by improving overall utilization of water and fertilizers needed to grow the water-thirsty almond nuts. Tom also wants to access real-time information about the state of his farm that he may use in making decisions to achieve the desired cost reductions. Tom’s problem is a perfect example for using a so-called smart farming application, an instance of a growing number of applications spanning domains as diverse as agriculture, transportation (asset management), housing (smart buildings), and infrastructure (smart power grids). What is common to these applications is the increasing use of the Internet of Things (IoT) to monitor operations in order to achieve the goal of reducing overall operating costs and making accessible real-time data about the state of the monitored entities. IoT is a paradigm of networking where the networked points (called “things”) are uniquely identifiable devices equipped with embedded connectivity. To achieve his stated goals, Tom installed wireless sensors throughout his nut plantation to measure the aforementioned variables. The measurements are then periodically sent to dedicated third party cloud-servers which compute the appropriate levels of water and fertilizer that should be applied to the plantation. Finally, the results are used to automatically drive the farm’s irrigation system by delivering precise amounts of water and fertilizers to the nut groves. The data archived on the cloud database is also processed by analytics applications. Our proposed research focuses on the design and implementation of storage, update, and transaction models for managing IoT data that flows from sensors to a cloud gateway, to a cloud database, to applications, and back. In the short term, we will investigate the architectural choices for the management of IoT data, and study data management (e.g., stream cleaning and fusion, and complex event detection) at the cloud gateway level by considering the IoT data requirements of gigantic volume, velocity and variety. In a longer term, we will study architectures for storage models of the cloud IoT distributed databases, as well as appropriate update and transactions models for IoT data management, and develop a prototype system. We anticipate that this research, which will enjoy the collaboration of a fortune 500 company as well as several renowned database and machine learning experts, will be significant with respect to new directions in the database industry, to IoT applications, as well as to the improvement of the IoT technology.