In P2P (peer-to-peer) streaming systems, server bandwidth consumption can be reduced by enhancing utilization ratio of nodes’ (users’) output (uplink) bandwidth capacity. With the ability of achieving maximum throughput of multicast, network coding has the potential to contribute to the enhancement. This article applies random linear network coding (RLNC) to P2P streaming system, and modeled transmission of P2P streaming. Greedy, rarest-first and random streaming algorithms are studied comparatively through streaming algorithm optimizations based on the framework of transmission model. Optimization results indicate that the random streaming algorithm that fetches data packets evenly and equally can utilize nodes’ output bandwidth more efficiently, which can reduce operating costs of service provider. Finally, by analyzing solutions of optimization model, guidelines are proposed as principles of streaming algorithm design for real P2P streaming systems.