A critical problem in a distributed system is the load of each of its individual components, which, if too large, threatens the availability of the whole system. A solution for system availability is decentralizing its components by using peer-to-peer networks. For such networks, we define in this work a scalable, transparent strategy for locally balancing the load of the individual peers, based on the intrinsic dynamicity of such networks, thus impacting as little as possible the availability and functionality of the system as a whole. We perform preliminary evaluation for our solution, too, concluding that the overhead added by our load balancing distributed algorithm is small and acceptable, considering the availability advantages it brings with it.