Video streaming services need a server selection algorithm that allocates efficiently network and server resources. Solving this optimization problem requires information about current resources. A video streaming system that relies entirely on the service provider for this task needs an expensive monitoring infrastructure. In this paper, we consider a two-phase approach that reduces the monitoring requirements by involving the clients in the selection process: the provider recommends several servers based on limited information about the system’s resources, and the clients make the final decision, using information obtained by interacting with these servers. We implemented these selection methods in a simulator and compared their performance. The results show that the two-phase selection is effective, improving substantially the performance of lightweight service providers, with limited monitoring capabilities.