Peer-to-peer system s have grown significantl y in popularity over the last few years. An increasing num ber of research projects have been closely following this trend, looking at many of the paradigm ’s technical aspects. In the context of data-sharing services, efforts have focused on a variety of issues from object location and routing to fair sharing and peer lif espans. Overall, the m ajority of these projects hav e concentrated o n either the whole P2P infrastructure or the clien t-side of peers. Little a ttention has been given to the peer’s server -side, even when that s ide determ ines m uch of the everyday user’s experience. In this paper, we m ake the cas e for looking at the serv er side of peers, focusing on the problem of sche dulin g with the in tent of minimizing the average respo nse time experienced by users. W e start b y charac terizing server workload based on extensive trace collection and analysis. W e then evalua te the perform ance and fairness of different scheduling policies through tr ace-driven sim ulations. Our results show that average response tim e can be dram atically reduced by more effectively scheduling the requests on the serv er-side of P2P system s.