Development of an Analysis tool for execution traces
Anders Johnsson, Roy Nilsson, Mälardalens högskola, Mälardalens högskola · 2004
In this thesis we present the Probabilistic Property Language (PPL) and a supporting tool. The purpose of this language is to analyse real-time systems based on information in execution traces. Traditional real-time analysis, e.g. fixed priority analysis, tends to be too pessimistic. Systems that does work can sometimes be deemed unschedulable since only worst case times, e.g. execution times or minimum inter arrival times, are considered. This is a problem when the worst cases are unlikely or never occur in the system. Furthermore these methods of analysis have no mean of checking properties like non emptiness of a message queue. To deal with that a simulation based analysis method was developed [13][14] . Using a modelling language a model of the system is constructed. Changes can then be made in this model before they are implemented in the system. Executing this model in a simulator will result in a trace. The trace contains information about task switches and probe observations that will be used to analyse the system with the help of PPL. A PPL query is formulated as a probabilistic statement, stating for example that some property should always be true, e.g. meeting a hard deadline, would be to formulate it as having a probability of 1. The probe observations contain information about system resource usage and other more application specific information. They could for example contain the size of a message queue. This could be used toverify requirements that could not be checked using fixed priority analysis like for example a message queue never being empty. Many different properties like precedence, separation and pre-emption can be formulated using PPL. In addition to this the query can contain an unbounded variable. This variable can be used to retrieve ...