Next-generation networks are currently being shaped by the softwarization of service components and the virtualization of resources. New approaches are being developed to control the compute, storage, and networking resources and to create, compose, and orchestrate different application components and network functions for offering services. The resource demands and the topology of virtualized composed services are, however, still fixed and pre-defined using rigid and inaccurate descriptors, usually created manually in current approaches. This jeopardizes the correctness of decisions for resource management and service orchestration and can easily result in over- or under-utilization of resources. The aim of this dissertation is to address this issue by introducing virtualized composed services, which have flexible structures and load-adaptive resource demands. In particular, different types of pliable virtualized composed services are described that consist of components, which (i) can be defined with a partial order and can be composed in different ways according to the availability of resources and the load, (ii) have their resource demands specified as a function of load and can be scaled horizontally and vertically depending on the load, or (iii) are developed in different deployment versions, each using different sets of virtual and physical resources, resulting in different characteristics, e.g., in terms of cost and performance. For pliable virtualized composed services, scaling, placement, and routing approaches are presented. These approaches can be used in modern service management and orchestration frameworks, to adapt the resource allocation and the topology of services to the requirements of service users, service providers, and network operators. Simulation-based analyses show the feasibility of defining services in a flexible and adaptable way.