Automatic computing resource awareness in resource managers for cognitive radios

Abstract

With all the potential flexibility of software-defined radios, the flexibility of SDR terminals is currently limited to design time flexibility. The capacity of the platform in terms of processing resources and internal bandwidths is dimensioned for the range of supported functionalities. In a platform-independent design scenario, resource managers play the role of matching waveform demands with platform capabilities. Predicting the task execution times has been studied in grid and distributed computing contexts with different objectives and assumptions. Given the dynamic nature of waveform demands as a function of the radio environment, an accurate characterization of the consumed resources can increase the efficiency of resource management strategies. In the SDR context, this efficiency translates to less energy consumption and higher resource utilization. Based on our experience acquired during the development of an SDR execution environment, this work presents the metrics that are needed by computing resource managers.