—The Session Initiation Protocol (SIP) is widely used for VoIP communication. Losses caused by network or server overload would cause retransmissions and delays in the session establishment and would hence reduce the perceived service quality of the users. In order to be able to take counter measures network and service planers require detailed models that would allow them to predict such effects in advance. This paper presents a theoretical model of SIP that can be used for determining various parameters such as the delay and the number of messages required for establishing a SIP session when taking losses and delays into account. The model is restricted to the case when SIP is transported over UDP. The theoretical results are then verified using measurements.