Research Report RR-14-296
Cognitive Networks have been proposed to opportunistically discover
and exploit licensed spectrum bands, in which the secondary users' (SU)
activity is subordinated to primary users (PU). Depending on the nature of
the interaction between the SU and PU, there are two frequently encountered
types of spectrum access: underlay and interweave. While a lot of research
effort has been devoted to each mode, there is no clear consensus aboutwhich
type of access performs better in different scenarios and for differentmetrics.
To this end, in this paper we approach this question analytically, and provide
closed-form expressions that allow one to compare the performance of the
two types of access under a common network setup. We focus on two key
metrics, delay and throughput, which we analyze using queueing theory and
renewal-reward theory. This allows an SU to decide when one type of access
technique or the other would provide better performance, as a function of the
metric of interest and key network parameters. What is more, based on this
analysis, we propose dynamic (hybrid) policies, that can decide at any point
to switch from the one type of access to the other, offering up to another 50%
of additional performance improvement, compared to the optimal "static"
policy in the scenario at hand. We provide extensive validation results using
a wide range of realistic simulation scenarios.