Spectrum sharing performance for multiplexed renewal and Poisson sources

The problem of sharing RF spectrum between two groups of users referred to as primary and secondary users is investigated in this thesis. Primary users are assumed to be licensed holders of spectrum that is under-utilized. The problem of spectrum sharing is motivated by the need to both accommodate secondary users that may operate in more congested regions of the spectrum and to better utilize the spectrum. This work investigates the performance degradation that primary users experience when secondary users are allowed to share the spectrum. The spectrum sharing system is analyzed using a queueing model that admits primary and secondary access to a finite number of channels. The blocking performance is derived assuming a first-in-first-out service model for the multiplexed arrival process generated by the primary and secondary arrivals. The primary users are modeled as Poisson arrival processes whereas the secondary users can be attributed a general renewal process to characterize their inter-access times. A computational algorithm is proposed for estimating the channel occupancy probabilities at the arrival instants, under the assumption that the multiplexed arrival process exhibits renewal properties. A simulation analysis is performed considering both renewal and exact arrival processes to validate the computational model and examine the effects of the renewal assumption. Gamma distributed inter-arrival times are considered for secondary arrival processes. The performance analysis considers the impact of changing index of dispersion of the renewal process and the number of excess channels available for spectrum sharing. It is found that by decreasing the index of dispersion of the secondary arrival process below the value of unity, the excess blocking experienced by the primary users can be reduced for the same offered loads. The performance can also be significantly improved as the number of excess channels increase. It is shown that neglecting the correlation in the multiplexed arrival process can lead to an under-estimation of the blocking probabilities of the primary users. The computational approach can however lead to a quick estimation of the performance metrics and with knowledge of the error bounds, is suitable for application towards admission control decisions that are required in dynamic spectrum allocation problems.