Scenario-aware workload characterization based on a max-plus linear representation

Abstract

This paper describes an event analytical model for the workload characterization of multimedia applications with several scenarios of operation. This model describes the application tasks as timed actors of a Scenario-Aware Dataflow Graph (SADF), so that the multiple application scenarios are defined in terms of the variable execution times previously identified in the timing characterization of every application task. A Max-Plus linear representation between the input and output event sequences of the SADF Graph states the basis of a model of scenario-based workload curves and a model of scenario-based service curves that allow to characterize the behavioral dynamism of the application in its design phase. By a detailed study, we show the applicability of our model for the performance evaluation of a multimedia application, whose tasks are implemented in general purpose processors.