In the recent years one of the most active research topics in multimedia networking is the exploitation of the Scalable Video Coding (SVC) as a scalable solution for efficient network resources utilization. SVC introduces scalability by exploiting a layered encoding of the video stream, thus enabling real-time in-network adaptation by selectively allowing the transmission of appropriate layers. This paper presents an architecture that exploits SVC capabilities in order to provide end-to-end QoS assurance via in-network video adaptation. The adaptation system management is based on MPEG-21 framework while the network QoS mechanisms are based on DiffServ standard. The performance evaluation of the proposed architecture is performed over a real test-bed infrastructure.