A Novel Scalable Cloud-enabled Spectrum Sensing Architecture

Spectrum Sensing techniques have evolved significantly, transitioning from traditional approaches to more complex Machine Learning and Deep Learning based approaches. This evolution has led to enhanced performance capabilities but has also introduced substantial computational demands, which in turn have constrained the scalability of large-scale deployments due to associated costs. Nevertheless, the advent of Cloud Computing offers a promising solution by enabling the centralized processing of data at the cloud level, thereby reducing the computational load on field-level sensors. This shift facilitates cost-effective deployment and enhances the scalability of spectrum sensing applications. In this study, we introduce and evaluate a Cloud-Enabled Spectrum Sensing architecture, with a specific focus on the computational and data transmission requirements from the perspective of the sensing node. We explore and compare two data formats for radio signal processing: I/Q samples and FFT series. Our evaluation covers the computational demands for processing and transmitting these data formats to a Cloud Data Fusion Center across a range of radio spectrum bandwidths. Moreover, we examine the latencies involved in data transfer. Our analysis reveals that the FFT series data format offers considerable advantages for this architecture, achieving an optimal balance between computational requirements and network load. Specifically, for the recording and transmission of a 30 MHz wide radio band using the TCP protocol, the FFT format requires only 0.3 GFLOPS for processing and 6.7 MB/s for data transfer. These findings underscore the potential of the FFT series as a highly efficient radio signal representation format for Cloud-Enabled Spectrum Sensing architectures, promising significant improvements in scalability and operational efficiency.