Peer-to-peer (P2P) streaming has become a promising approach for disseminating streaming media content from the server to a large number of interested clients. It still faces many challenges, however, such as high churn rate of peer clients, uplink bandwidth constraints of participating peers, and heterogeneity of client capacities. To resolve these issues, this paper presents a new P2P streaming framework that combines with the advantages of both mesh-based multisource overlay networks and scalable video coding techniques, specifically with multiple description coding (MDC), to improve the streaming quality of participating clients. In this paper, an optimized allocation policy (OAP) algorithm was proposed for multidescriptions allocation. Extensive simulations show that the proposed system achieves higher quality of service by peer-assisted cooperative streaming and MDC coding. In addition, we investigate an efficient cooperative caching mechanism for streaming service. The target is to provide low-latency and high-quality services by virtue of peer collaboration. The storage and replacement of caching content are based on a segment-based strategy. Through comparison, we demonstrate the effectiveness of the proposed scheme and compare with the traditional LRUF scheme through extensive experiments over large Internet-like topologies. Results show that the system outperforms some previous schemes in resource utilization and is more robust and resilient to node departure, which demonstrate that it is well suited for quality adaptive cooperative streaming applications.