Femto base stations will be widely deployed in 5G times for compensating poor cellular coverage for use in residential or small business environments. In this paper, we propose a spectrum sharing framework, where a cluster of femtocells are under the control of a centralized local manager. The framework assumes that each femto base station has the sniffing ability that can detect its neighbor femtocells and the controller can then have complete knowledge of local interference constraints. Unlike distributed WiFi that runs on unlicensed band, the femtocell owned by a carrier runs on licensed band and the carrier prefers a centralized system to a distributed one. Therefore, a fair spectrum sharing policy on the licensed band is more important among carriers to break the monopoly and inefficient spectrum allocation. Considering the spectrum can be spatial reused, an interference graph is adopted to avoid the interference between femtocells of all the carriers. We test the framework not only by comparing the average number of channels per user can get but also the variance to guarantee the fairness. In this framework, the scheme combining topological sort and maximum traffic load of femtocell contributes to keep a better balance on the average number of channels and variance.