A Testbed for Hardware-assisted Online Profiling of IoT devices

Abstract

The widespread application of the Internet of Things (IoT) has put forward higher requirements for the reliability of the IoT devices. Traditional testing methods, while able to get a rough approximation of the performance of IoT devices, often fail to extract detailed runtime execution traces of applications from the resource-constrained devices. Hardware-assisted tracing can make it easier for IoT developers to obtain the rich running information of IoT devices with limited overhead, which brings new possibility to fully evaluate the software (i.e. firmware) of the IoT. For the hardware-assisted tracing data, existing offline analysis methods have severely limited the observable time span due to a large amount of tracing data. In this paper, we propose an FPGA-based online profiling testbed to carry out real-time processing of tracing data and implement continuous observation of IoT devices. Our preliminary experiment shows that the testbed has superior performance in terms of runtime trace capturing and sampling frequency. Finally, the potential applications of the captured traces are discussed.