Demultiplexing Considerations for Statistical Multiplexors

Demultiplexing serves as an important function for statistical multiplexors. Its purpose is to reassemble the received message and distribute it to the appropriate destination. An important cost consideration for this function is the size of the buffer necessary to meet a specified overflow service requirement. The demultiplexing buffer can be modeled as a finite waiting room queueing model with batch Poisson arrivals and multiple distinct constant servers. Stimulation is used to study buffer behavior for traffic arriving at the buffer according to the uniform, linear, step, and geometric destination functions. The relationships among buffer overflow probability, buffer size, traffic intensity, average message length, and message destination are presented in graphs to provide a guide in the design of demultiplexing buffers. Simulation results reveal that buffer input messages that have short average message lengths and uniform traffic destinations yield the best buffer behavior. Thus, in planning the CPU scheduling algorithm and in selecting the demultiplexing output rates, the designing of a computer communications system that uses the statistical multiplexing technique should also consider the output statistics needed to achieve optimal demultiplexing performance.