In this paper we propose a model for a service facility which caters to two types of customers. One motivation for this model is a trunk group which is used simultaneously for voice and data communication. We assume that primary customers receive preemptive priority service and leave the system without service if on arrival they find all servers busy with other primary customers. The facility has a queueing provision (buffer) for secondary customers who are served only when the facility is not fully occupied by primary customers. The motivation for this model is discussed. Numerical examples of the statistics of the content of the buffer are presented.