The hybrid traffic, a composite of elastic and inelastic data generated by delay-tolerant and real-time applications respectively, forms the bulk of what is carried over modern cellular systems. Since a smooth performance of realtime and delay-tolerant applications is tied to their distinctive needs for resources, networks' intelligence about representative characteristics of applications' traffic, networks' ability to prioritize subscribers, and networks' sophistication to cope with temporal variations in traffic amounts from the real-time and delay-tolerant applications can help efficiently assign resources accordingly to the preceding issues and consequentially elevate the quality-of-experience. To this end, the manuscript at hand aims at concocting a convex proportional fairness resource allocation method for cognitive cellular systems equipped with the intelligence to consider the traffic type, its temporal variations, and user prioritization. The devised method is formalized as mathematically equivalent centralized and distributed formulations with provided solution algorithms and convergence conditions. It is proved that the centralized approach has a lower transmission overhead, whose some lower bounds are derived for the centralized and distributed methods. Analyses of the sensitivity to changes of user equipments quantity and of application usage are presented under a variation of bidding scenarios in the face of the system dynamics.