Scenario-based prediction of run-time resource consumption in component-based software systems

Abstract

Resources of embedded systems, such as memory size and CPU power, are expensive and (usually) not extensible during the lifetime of a system. It is therefore desirable to be able to determine the resource consumption of an application as early as possible in the design phase. Only then, a designer is able to guarantee that an application will fit on a target device. Resource prediction is a technique to estimate theamount of consumed resources by analyzing the design and/or implementation of an application. In this paper we concentrate on predicting memory consumption in component-based applications. Component-based applications complicate resource predictions because resource consumption is spread across individual components. The challenge is to express resource consumption per component, and to combine them to do predictions over compositions of components. To that end, we propose a model in which individual resource estimations of components can be combined. These composed resource estimations are then used in scenarios (which model run-time behavior) to predict memory consumption of applications.