In this paper, we propose and analyze a TS (Tabu Search) algorithm for DSA (Dynamic Spectrum Access) in cellular networks. We consider a scenario where cellular operators share a common access band, and we focus on the strategy of one operator providing packet services to the end-users. We consider a soft interference requirement for the algorithm's design that suits the packet traffic context. The operator's objective is to maximize its reward while taking into account the trade-off between the spectrum cost and the revenues obtained from endusers. We focus on the temporal heterogeneity of the traffic and we show that our algorithm allows the operator to increase its reward by taking advantage of this heterogeneity, rather than assuming homogeneous traffic for spectrum allocation. We study the dynamicity of the algorithm through event-based simulations. Results show that our algorithm uses less spectrum and achieves less blocking probability than the FSA (Fixed Spectrum Access) case at the price of reduced user throughput.