Network virtualization has been propounded as a key attribute of the future internetworking paradigm and is expected to play a crucial role in the next generation Internet. In network virtualization environments, service providers utilize the networking resources provided by multiple network infrastructures to offer end-to-end services. Therefore, resource allocation for end-to-end Quality of Service (QoS) provisioning in network virtualization becomes a significant research problem. This paper investigates this problem from a service provider's perspective. The main contributions of this paper include an analytical model for end-to-end service provisioning in network virtualization and the analysis techniques for allocating resources in network infrastructures to provide end-to-end QoS guarantees. The modeling and analysis techniques developed in this paper are general and flexible; thus are applicable to the various heterogeneous networking systems coexisting in the future Internet.