Network virtualization is an important concept to overcome the ossification of today's Internet as it facilitates innovation also in the network core and as it promises a more efficient use of the given resources and infrastructure. Virtual networks (VNets) provide an abstraction of the physical network: multiple VNets may cohabit the same physical network, but can be based on completely different protocol stacks (also beyond IP). One of the main challenges in network virtualization is the efficient admission control and embedding of VNets. The demand for virtual networks (e.g., for a video conference) can be hard to predict, and once the request is accepted, the specification / QoS guarantees must be ensured throughout the VNet's lifetime. This requires an admission control algorithm which only selects high-benefit VNets in times of scarce resources, and an embedding algorithm which realizes the VNet in such a way that the likelihood that future requests can be embedded as well is maximized. This article describes a generic algorithm for the online VNet embedding problem which does not rely on any knowledge of the future VNet requests but whose performance is competitive to an optimal offline algorithm that has complete knowledge of the request sequence in advance: the so-called competitive ratio is, loosely speaking, logarithmic in the sum of the resources. Our algorithm is generic in the sense that it supports multiple traffic models, multiple routing models, and even allows for nonuniform benefits and durations of VNet requests.