Satellite network is regarded as an important component of 5th Generation (5G). New deployment scenarios have been designed in form of satellite network fused 5G terrestrial network. Service requests in Non terrestrial networks (NTN) are non-centralized in the time and space domain, which cause services modelling uncertain and lack of balance between latency and bandwidth. Targeted at those bottlenecks, a Satellite service model based on State machine with tracking tree (SM-SMTT) and traffc allocation scheme base on cumulative entropy is proposed. Enhanced mobile broadband service, low latency service and massive machine type communications in satellite will be modelled by SM-SMTT. Cumulative entropy is introduced as the criterion to allocate traffc. According to the lower bound from cumulative entropy, the balance of latency and bandwidth is achieved in different deployment scenarios. The simulation in OPNET, demonstrations the new scheme has promoted the performance of delay and bandwidth utilization by 37.8％ and 12.0％.