Mobile Edge Computing (MEC) emerged as an alternative to reduce network latency by bringing the data processing close to the users. This proximity requires that the service is available most of the time. Trying to assess the availability of such systems in real contexts requires high costs. This paper uses Stochastic Petri Nets (SPNs) to assess the availability of an MEC architecture, seeking to avoid premature investment in real equipment. In addition, this work presents a sensitivity analysis that identifies the most critical components of architecture. This work has the potential to assist administrators in redefining the most optimized MEC architectures by decreasing the risk of failure.