With the trend of platformization and the advent of multicore processors, we need to extend the traditional resource modeling in Service-Oriented Infrastructure to enable a better adoption of platform-level features such as multicores. For service processes that are run on multiple servers, each on a multicore architecture, an end-to-end approach should be used to ensure a consistent QoS from all service components. In this paper, we propose an end-to-end QoS management architecture for coordinating the performances of service components. Our research considers modern performance enhancing technologies, namely, multicore and virtualization, and proposes an angel thread mechanism to manage service QoS. We also study scheduling approaches for the architecture.