This paper presents VoIP reliability design methods and tools that have been used successfully in our work. It is based on the standards-driven architectural requirements for an enterprise VoIP network. Lucent's Accelerate Enterprise Solution (AES) is chosen as a particular example of a flexible enterprise VoIP network that is used both as an in-house enterprise VoIP solution and as an hosted IP telephony service. This paper discusses an enhanced method and procedure of the reliability calculation, using a network matrix-representation. This approach allows for easier modeling setup and calculations as well as offering quick discussion on design alternatives with customers. The modeling methodology is characterized by its flexibility to fit different architectures, i.e. from a simple reference-connection to a complex multi-path network. Typical VoIP requirements and benchmarks are discussed and compared to the modeling results. The observations highlight the challenges that a typical enterprise VoIP network encounters. This reliability modeling work is an integral part of the reliability service level agreement (SLA) framework. A discussion of the risk for the design of SLA is also presented. The paper concludes by discussing the impact of network security on the network availability, which is an area where industry needs to pay attention