Digital Twins are a key Industry 4.0 application, providing intelligent, real-time replicas of physical assets such as industrial robots so they can be controlled remotely. The study deploys an Edge-based Digital Twin for remote control of robotic arms in an experimental testbed, splitting the service into virtual network functions following the NFV paradigm. It evaluates how overall service performance relates to the resource needs of each VNF and to the number of robots, and how the abstraction level of the commands sent to the robots creates a trade-off between computing resources and trajectory guarantees. The results inform the design of scaling and orchestration frameworks for Digital-Twin-as-a-Service, and the experiments show that LTE cannot meet the minimum latency a Digital Twin requires.