Edge Computing for IoT Sensors Based on DPN
Tomoaki Kanaya, Akihiro Nakao, Shu Yamamoto, Masato Oguchi, Saneyasu Yamaguchi · 2020
IoT (Internet of Things) technology enables us to gather a huge amount of data from many sensors in various places. In case every IoT sensor directly sends its data to the server, the load of the server must be too large. Several ideas have been proposed for addressing this issue. Data consolidation at edge computer, which is called edge-computing, is one of the most promising methods. In the cases of the existing methods based on edge computing, an IoT sensor always connects its edge computer. Thus, data are always processed in the edge server and the place for processing cannot be dynamically optimized based on loads of edge and centralized servers. In this paper, we propose a novel transparent edge computing based on DPN (Deeply Programmable Network) for non-encrypted IoT data uploading. The programmable switch deeply inspects the payloads of non-encrypted packets with the proposed method. The switch then processes the data and decreases the number of packets that are transmitted to the server. We introduce our prototype system for calculating top-K and show our method works effectively.