The Impact of Frequency-Agility on Dynamic Spectrum Sharing

Designed to adapt spectrum usage on-the-fly, frequency-agile radios can drastically improve performance of wireless networks. Such flexibility, however, comes with a cost of increased hardware complexity. This motivates us to understand when and why having higher degree of frequency-agility helps and how much improvement it can lead to. In this paper, we approach this question by comparing two types of agile radios in the context of dynamic spectrum sharing in any given spectrum chunk. We consider 1-agile radios that use a single frequency channel but can adjust the channel's width and central frequency, and k-agile radios that can combine up to k non-contiguously aligned frequency segments into one transmission. We show that, due to inherent demand dynamics and conflict heterogeneity, networks using 1-agile radios often face the problem of spectrum fragmentation. But k-agile radios can effectively suppress this problem directly at the physical layer. Using theoretical analysis and simulation experiments, we quantify the advantage of k- agile radios over 1-agile radios in their network spectrum usage. For a fair comparison, we abstract the impact of demand and topology configurations by evaluating the worst case and average case performance. Our results show that in worst cases, the improvement of using fully-agile radios is arbitrarily large, although the improvement of using k-agile radios is upper bounded by k. In average cases, the improvement reduces to 10-40% under typical network configurations. Interestingly, in the context of dynamic spectrum sharing, 2-agile radios realize the majority of the improvement brought by fully-agile radios.