Performance analysis for embedded software design

Hettinga, S. (2010) Performance analysis for embedded software design.

[img]
Preview
PDF
1MB
Abstract:to reduce the number of implementation errors in the integration phase of a software project, new software development methods are required. The Analytical Software Design (ASD) method of the company Verum provides design strategies combined with model-checking to reduce the number of errors made when developing software. Currently there is no insight into the performance of a software system, before the implementation phase of a project has been finished. When a system does not meet the performance requirements after integration, improving the software speed in this stage of the project is a very expensive and time consuming process. To provide early stage insight into the performance of a software system, adequate performance models and corresponding analysis tools are needed. In the thesis a start is made by modelling and analysing the performance of the software generated by the ASD suite. A hierarchical approach is used to model the whole ASD generated software. Queuing theory is applied to model the blocks, that are generated by the ASD suite. Waiting time propagation then composes the model for the whole system from the individual blocks. The focus of this thesis is on making accurate models, with scalable analysis techniques. Single ASD blocks are modelled as queueing stations and a case study shows that the analysis results match the simulation results accurately. To model the dependencies between different blocks, phase type distributions are used. Also systems composed of multiple blocks are analysed and validated using simulation. The analytic results for larger systems deviate substantially from the simulation, because some dependencies are left out of the model. Recommendations are given to improve the quality of the results in future research. One way to do this, could be tto embedded these dependencies into the model by also using phase type distributions for elements that are currently modelled as negative exponential distributions.
Item Type:Essay (Master)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:54 computer science
Programme:Embedded Systems MSc (60331)
Link to this item:http://purl.utwente.nl/essays/69539
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page