In this paper, the authors investigate capacity optimization mechanisms for multi-beam satellite systems built on a realistic payload model. The first proposed mechanism deals with long term traffic variations, for which capacity optimization algorithms are proposed based on per-beam traffic requests. Due to the high asymmetry of the traffic, their algorithms provide time and spatial flexibility illuminating a specific set of beams within a window of several time-slots. Their algorithms maximize the amount of capacity actually offered while providing reduced power consumption. The second proposed mechanism deals with short-term traffic variations, for which they propose Network Coding (NC) based techniques at the link layer.