FIFO Scheduling and Event Count Curves for Modeling Structured Event Streams in Modular Performance Analysis

Abstract

The growing complexity of distributed embedded real-time systems requires elaborate models and methods in their design process. While simulation does not sufficiently cover corner cases and may have excessive run times for complex systems, analytic methods for system level performance analysis have been established in the past. Their efficiency in computing hard bounds on buffer sizes, end-to-end delays or throughput has proven their usefulness. One of the major drawbacks of these methods is their limited scope in terms of system classes that can be analyzed with high accuracy. In this paper we extend existing methods for analyzing heterogeneous distributed embedded systems such that different types of data streams can be composed to a higher level event stream with multiple hierarchies. The method is based on a novel characterization of structured event streams, i.e., event streams for which the individual events belong to a finite number of classes. The new class of methods we provide can be embedded into well known compositional frameworks for performance analysis such as Symta/S and MPA (Modular Performance Analysis). We propose and compare two different approaches which take advantage of event stream structures. Realistic examples are given that are used to apply these newmodels and methods to performance analysis.