Active-network technology envisions deployment of virtual execution environments within network elements, such as switches and routers. As a result, application-specific processing can be applied to network traffic. To use such technology safely and efficiently, individual nodes must provide mechanisms to manage resource use. This implies that each node must understand the varying resource demands associated with specific network traffic. Well-accepted metrics exist for expressing bandwidth (bits per second) and memory (bytes) in units independent of the capabilities of particular nodes. Unfortunately, no well-accepted metric exists to express processing (i.e., CPU time) demands in a platform-independent form. This paper describes and evaluates an approach to model the CPU demand of active packets in a form that can be interpreted among heterogeneous nodes in an active network. The paper applies the model in two applications: (1) controlling CPU use and (2) predicting CPU demand in an active-network node. The model yields improved performance when compared against the approach currently used in many active-network execution environments. The paper also discusses the limits of the proposed model, and outlines future research that might lead to improved outcomes.