In every production and/or inventory control operation, the reliability of the supply process has a great impact on its outcome. We present two analytical models for computing the performance of a merge, push supply network with unreliable suppliers that feed a distribution centre with a common shared buffer. The breakdowns and repair times of the suppliers are modelled via exponential and Coxian-2 distributions. The supply network is modelled as a continuous time Markov process with discrete states. The structure of the transition matrices of these specific systems is explored and two separate computational algorithms are developed to generate them for different values of system characteristics. The proposed algorithms are used as an optimisation tool to determine the optimal values of the system's characteristics such as the number of suppliers, capacity of distribution centre, capacity of buffer and to calculate the system's performance measures. Also a comparison between the two proposed algorithms is provided regarding the performance measures of WIP system , throughput rate, availability and blocking of each supplier.