Skip to Main Content
Developments toward ubiquitous and pervasive computing have lead to application scenarios where a large number of sensors and computing devices are connected to the network. These devices might constantly send status information via multicast to a number of applications or users. One big challenge in this environment is the amount of data traffic generated by such sensors, which depends on the size of the data, the transmission frequency, and the number of senders and receivers. However, for certain applications it is sufficient to receive less accurate, aggregated data from a group of sources. This leads to the possibility of using programmable routers to perform such data aggregation inside the network. While the basic algorithm for data merging has been described in literature, we address how a large number of sources can be organized in a hierarchical structure to allow multiple users to get a view of all sensors at different levels of aggregation. With the control information that is provided by an aggregating node, the user can traverse the aggregation tree by joining different multicast sessions that transmit different levels of detail. This provides a novel communication paradigm that reduces the network overhead from continuously transmitting sources and organizes data in a way that it is useful to receivers. We provide two detailed example scenarios: A battlefield information system, which aggregates geographic location data of units, and a conferencing application, which aggregates audio data. We describe the aggregation algorithm that is used and analyze its effect on delay and jitter of periodic transmissions by the sources. We describe the hierarchical control structure that provides multiple levels of aggregation and how real-time transport protocol (RTP) can be used to implement it. The performance of the proposed scheme is evaluated with measurements that were done on an implementation of the audio aggregation application.