To improve the quality of service of web caching system and reduce user perception time, a Scalable Performance-Constrained Modeling (SPCM) for Institution-based Peer-to-Peer Web Caching System is proposed in this paper. Performance-Constrained means that the total time of object orientation and access is smaller than that from origin server directly. SPCM system distinguishes itself from others with this feature. Using trace-driven simulation, we compared and evaluated the latency improvement of all design choices. The experiments show that SPCM can improve the hit ratio, reduce the access web object latency.