Automatically improve software architecture models for performance, reliability, and cost using evolutionary algorithms

Abstract

Quantitative prediction of quality properties (i.e. extrafunctional properties such as performance, reliability, and cost) of software architectures during design supports a systematic software engineering approach. Designing architectures that exhibit a good trade-off between multiple quality criteria is hard, because even after a functional design has been created, many remaining degrees of freedom in the software architecture span a large, discontinuous design space. In current practice, software architects try to find solutions manually, which is time-consuming, can be errorprone and can lead to suboptimal designs. We propose an automated approach to search the design space for good solutions. Starting with a given initial architectural model, the approach iteratively modifies and evaluates architectural models. Our approach applies a multi-criteria genetic algorithm to software architectures modelled with the Palladio Component Model. It supports quantitative performance, reliability, and cost prediction and can be extended to other quantitative quality criteria of software architectures. We validate the applicability of our approach by applying it to an architecture model of a component-based business information system and analyse its quality criteria trade-offs by automatically investigating more than 1200 alternative design candidates.

Research Preview: Prioritizing Quality Requirements Based on Software Architecture Evaluation Feedback

“…PerOpteryx, ArchE, Performance Booster, Archeopteryx [11]). Such approaches automatically vary a given architectural model based on predefined degrees of freedom, such as component allocation to servers, component selection, change of hardware and software parameters, or other, custom defined design decisions expressed as simple model transformations.…”

“…Only after initial architecture evaluation and design space exploration, the trade-off between quality attributes and the costs for achieving quality levels can be reliably estimated. To validate our research idea, we will (1) extend the existing design space exploration tool PerOpteryx [11] to explicitly support quality requirements prioritization and (2) evaluate its benefits in empirical studies, which include business reporting and industrial automation systems. The expected results of our approach are (1) better informed quality level definition, (2) guidance in quality requirement prioritization, and, as a result, (3) higher trust in quality requirements during the development process.…”

A Process to Effectively Identify “Guilty” Performance Antipatterns

“…In this side a list of existing alternatives can contain a number of options for what could be changed in the software system model. From this list, alternatives can be chosen to directly create a new software system models [9]. Alternatives to apply can be chosen randomly, manually by software architects based on their experience, or based on heuristics.…”

“…In another previous work we have proposed a complementary approach to improve software performance for component-based software systems based on metaheuristic search techniques [9]. We proposed to combine random moves, as shown on the right hand side of Figure 1, and heuristic rules to search the given design space.…”

QVTR2: A Rational and Performance-Aware Extension to the Relations Language

“…Existing approaches provide poor or no support at all to interpret results, to relate them from lowlevel abstraction layers to high-level layers, and to identify appropriate solutions when requirements are not met. Methodologies for feedback provision to engineers are already available in literature; example are the meta-heuristic approaches described in [3][4] [5] or the rule based approaches described in [6][7] [8]. Indeed, meta-heuristic approaches work only in particular contexts (e.g., component based engineering) and provide solutions only for specific performance issues (e.g., finding an optimal allocation for components).…”