DisDaS: Distributed Data Streams in Dynamic Environments

我们目前注意到，人们对大型设备系统的兴趣正在迅速增长，每个设备系统都会永久地观察数据，这些数据通常是实时的，这些数据将被聚合为有用的信息。这类系统的例子是(1)全球分布的用户的智能手机收集的信息可以(在某些情况下已经)用于聚合信息。除了由供应商进行监控外，还可以请求基于位置的信息，如附近的餐馆、前面的交通信息、附近的朋友和许多其他类型的信息。(2)汽车产生传感器数据(例如，关于它们的位置、速度、环境和附近的其他汽车)，用于实现在十字路口驾驶或在高速公路上通过的自组织管理。(3)互联网服务提供商的网络节点观察本地链路的使用情况，并共同工作，以保持网络的健康状态。(4)传感器或机器人被部署到某个领域，目的是从观测数据中收集有用的信息。聚合的例子有(加权的)测量温度的平均值、最小值、最大值。在这个项目中，我们计划为分布式算法的设计和分析奠定基础，这些算法可以连续计算大量设备观察到的数据流的聚合信息。这些设备可以是移动的，即能够在飞机或空间中移动，并包含(无线)通信设备和用于观察其环境的传感器。主要的挑战是处理这些设备产生的海量数据。通常，数据流太大，到达速度太快，无法完全存储，或通过网络发送到中央服务器，或实时处理。因此，我们必须设法利用有限的资源，如内存、通信量和计算时间，从数据流中提取有用的信息。我们计划在动态环境中开发连续分布式算法，同时考虑设备的移动性和观察到的事件的移动性。这反映了用智能手机观察环境的人的移动场景。一方面，设备的动态模型和数据流的生成将受到我们的理论模型的推动。此外，我们期待着与处理社交网络的优先计划大数据算法(SPP 1736)中的项目就此类模型进行有趣的合作。

We currently observe rapidly growing interest in large systems of devices, each of which permanently observes data that has - often in real time - to be aggregated to useful information. Examples for such systems are (1) Information gathered by the smartphones of world-wide distributed users can (and in some cases is already) used for aggregating information. Besides the monitoring done by the providers, also position based information like nearby restaurants, information about traffic ahead, nearby friends and many further kinds of information can be requested. (2) Cars generate sensor data (about, for example, their position, their speed, their environment, and other cars nearby) that is used in order to realise a self-organised management for driving in intersections or for passing on freeways. (3) Nodes of a network of an Internet Service Provider observe local usage of links, and work together in order to keep the the network in a healthy state. (4) Sensors or robots are deployed to some field, with the aim of aggregating useful information from observed data. Examples for aggregation are (weighted) average, minimum, maximum of measured temperature. In this project we plan to lay the foundations for the design and analysis of distributed algorithms that continuously compute aggregated information of streams of data which are observed by a multitude of devices. These devices may be mobile, i.e. capable of moving in the plane or in space, and contain both (wireless) communication devices and sensors for observing their environment. The major challenge is to cope with the huge amount of data generated by the devices. Typically, the data streams are too big and arrive too fast to be completely stored, or sent to a central server through a network, or processed in real time. Thus we have to find ways to extract useful information from the streams using restricted resources like memory, communication volume and computation time. We plan to develop continuous distributed algorithms in dynamic environments, taking both mobility of the devices and of the observed events into account. This reflects the scenario of moving people with smartphones who observe their environments. The models for the dynamics of the devices and the generation of the data streams will on one side be motivated by our theoretical models. Moreover, we expect interesting cooperation about such models with projects within the Priority Programme Algorithms for Big Data (SPP 1736), that deal with social networks.