The ability to dynamically configure mobile ad-hoc network (MANET) devices is critical for supporting complex services such as quality of service (QoS), security and access control in these networks. In our work, we address the problem of policy distribution and provisioning in MANETs. Previously, we have proposed a policy provisioning architecture that enables the operator, e.g., a military command and control center, to maintain a logically centralized control of the network as a whole, while allowing a physically decentralized and self-managing implementation essential for a MANET environment. In this paper, we present an analytical model of the availability of the policy distribution service in a MANET and analyze the performance using stochastic Petri nets (SPNs). We develop the model at two levels of complexity: as a simple Markovian model with Poisson assumptions and as a more accurate non-Markovian model with general distributions obtained using statistical parameterization. We compare and cross-validate the analytical results with simulation and experimental results. Finally, we illustrate the effectiveness of the architecture for managing QoS for soft real-time applications, using an emulated mobile ad-hoc network testbed.