Real-time performance modeling of link layer protocols for multi-layer protocol aggregation

Abstract

A hybrid modeling method for real-time performance evaluation of link layer protocols is presented which makes use of task graph structures, Petri Net synchronization elements, general stochastic arrival/service processes, and channel error characteristics. The resulting models are analyzed exactly by probabilistic task aggregations leading to a separation of the protocol and queuing functions by which the protocol model is stepwise reduced to an aggregated frame transit time representation acting as a virtual service time Tx of a standard queuing model GI/G/n. The methodology is applied to two classical communication protocols: (1) the Stop- and-Wait (SW) Protocol and (2) the Selective-Repeat (SR) Protocol, both with positive Acknowledgements and Timeout Recovery (ACK/TO). The method is applied to the performance analysis of multi-layer architectures to reduce complexity and increase accuracy. The method is demonstrated for Networked Control Systems (NCS) where the link layer delay is embedded within the control loop by an equivalent stochastic phase and where delay threshold percentiles have to be guaranteed. The method is also a key to the performance evaluation of multi-layer protocol architectures where a lower layer subsystem is aggregated into a stochastic phase which can be inserted in the next higher protocol layer, applied layer by layer repeatedly.