A Sensor-aware Offloading Model for IoT Edge Computing

Abstract

In IoT sensor network environment, offloading is an important factor that affects all design objectives. Since massive amounts of data are collected every second to the gateway and so immediate processing is difficult, offloading is critical to quickly eliminate worthless data in advance. Similar sensor data are continuously generated except in abnormal situations such as sudden changes and failure events. Therefore, the amount of data processing and frequency of data transmission can be greatly reduced by classifying, filtering, and compressing the data. In addition, more meaningful IoT context can be analyzed by combining multiple sensor data, since the sensor values generated by each sensor has its own characteristics. The previous offloading techniques mainly focused on minimizing latency without using data context and data resizing. Therefore, a new filtering technique is required to enhance the offloading efficiency through precision control using sensor context patterns. This paper proposes a new sensor-aware context offloading model called SCOM to support efficient data filtering services for the edge-based IoT environment. The architecture of SCOM consists of three layers of sensor context, pattern context and transmission context. SCOM exploits context-aware stream pattern matching using general string matching based on slide window for sensor stream offloading. Experiments show that the performance gain of SCOM reaches to 14.8% with respect to the operation throughput. Since proposed data layering and pattern-based offloading scheme can improve the sensor data filtering performance in edge gateways, it can be used for IoT sensor monitoring applications.