PAD-A: performance antipattern detector for AADL
- PDF / 666,960 Bytes
- 10 Pages / 595.276 x 790.866 pts Page_size
- 7 Downloads / 180 Views
ORIGINAL RESEARCH
PAD-A: performance antipattern detector for AADL Abrar Ul Haq Syed1 • Javaid Iqbal1
Received: 2 March 2020 / Accepted: 6 November 2020 Ó Bharati Vidyapeeth’s Institute of Computer Applications and Management 2020
Abstract In this article we have investigated the use of software performance antipatterns in tackling the performance issues in software architectures modelled using architecture analysis and design language (AADL). We have implemented our approach and developed a tool for OSATE (open source AADL tool environment) called PAD-A (performance antipattern detector for AADL). We evaluate our approach by applying it to two case studies. Results suggest effectiveness of antipatterns in tackling performance problems in early development phases. Keywords Software performance antipatterns OSATE plug-in AADL Architecture optimisation Architecture refactoring
1 Introduction Use of quantitative techniques to predict performance of software system early in design phase of software development life cycle allows software designers to identify the feasible design alternative and eliminate the poor ones well before implementation and thus saving lot of cost and efforts of rework that is involved in tuning at later stages of software development process. Performance of system architecture can be evaluated either using simulation methods or by analysis method or hybrid method i.e. & Javaid Iqbal [email protected] Abrar Ul Haq Syed [email protected] 1
University of Kashmir, Srinagar, India
combination of simulation and analysis methods [1]. In simulation, performance of a system is assessed by simulating all functions and characteristics of real running system. This approach is very difficult and cost and time intensive due to the complexity of software systems. In analysis method, performance indices are assessed using performance annotations provided in the software model. Analysis methods are cost efficient and provide quantitative results. Well-assessed techniques are available for prediction of software performance in early development phases using analysis method [2]. For unified modelling language (UML), the model is first annotated with performance parameters, which is then used to create a performance model (like Queuing Networks, PetriNets etc.). Performance indices are then predicted by quantitative analysis of these performance models. In this paper we have investigated architectural model, architecture analysis and design language (AADL), for performance improvement. Syntax and semantics of AADL allows us to specify performance parameters on AADL model and its supporting tools allow evaluating the architecture performance without converting them to performance models. AADL is used to model real time, performance critical, embedded and distributed computer systems [3] and is published as Society of Automotive Engineers (SAE) standard AS-5506 [4]. AADL provides the constructs to model hardware (execution platform) and software components as well as their bindings. Software co
Data Loading...
