Future network services and applications, such as coordinated remote driving or remote surgery, pose serious challenges on the underlying networks. In order to fulfill the extremely low latency requirement in combination with ultrahigh availability and reliability, we need novel approaches, for example to dynamically move network “capabilities” close to the users. This requires more flexibility, automation and adaptability to be added to the networks at different levels and operation planes. The key enabler of the novel features is network softwarization provided by NFV and SDN techniques. In this paper, we focus on a central component of the orchestration plane which is responsible for mapping the building blocks of services to available resources. Our main contribution is twofold. First, we propose a novel service graph embedding algorithm which is able to jointly control and optimize the usage of compute and network resources efficiently based on greedy heuristics. Besides, the algorithm can be configured extensively to obtain different optimization goals and trade-off running time with the search space. Second, we report on our implementation and integration with our proof-of-concept orchestration framework ESCAPE. Several experiments confirmed its practical applicability.