A computational method is developed to evaluate the data loss rate in a communication buffer with finite capacity. Both the producer ( sender) and the consumer (receiver) of this buffer are represented by semi-Markov processes. A data miss occurs when the producer attempts to add a new message to a full buffer. In practice, the capacity of the buffer must be carefully chosen to meet specified maximum data loss rates. The analysis presented can be used to optimize the design of communication systems by determining the appropriate buffer capacity for a required performance level. A semi-Markov model derived from actual measurements of a system program is examined. Useful analytical relations that can be used to compute or validate the parameters of this semi-Markov model are presented