Scaling Size and Parameter Spaces in Variability-Aware Software Performance Models (T)

Abstract

Abstract-In software performance engineering, what-if scenarios, architecture optimization, capacity planning, run-time adaptation, and uncertainty management of realistic models typically require the evaluation of many instances. Effective analysis is however hindered by two orthogonal sources of complexity. The first is the infamous problem of state space explosion-the analysis of a single model becomes intractable with its size. The second is due to massive parameter spaces to be explored, but such that computations cannot be reused across model instances. In this paper, we efficiently analyze many queuing models with the distinctive feature of more accurately capturing variability and uncertainty of execution rates by incorporating general (i.e., non-exponential) distributions. Applying product-line engineering methods, we consider a family of models generated by a core that evolves into concrete instances by applying simple delta operations affecting both the topology and the model's parameters. State explosion is tackled by turning to a scalable approximation based on ordinary differential equations. The entire model space is analyzed in a family-based fashion, i.e., at once using an efficient symbolic solution of a super-model that subsumes every concrete instance. Extensive numerical tests show that this is orders of magnitude faster than a naive instance-by-instance analysis.

An efficient method for uncertainty propagation in robust software performance estimation

“…For instance, Menasce et al [41] deal with uncertainties by reconfiguring software architectures at runtime. Kowal et al [42] leverage the commonalities across variants of software systems to analyse uncertainty in terms of both parametric changes (which affect the values of performance annotations) and structural changes (which may affect the topology of performance models). Incerto et al [43] express uncertainty as symbols in the specification of queuing network performance models and derive actual values of symbols by means of satisfiability modulo theory.…”

ERODE: A Tool for the Evaluation and Reduction of Ordinary Differential Equations

“…Ordinary differential equations (ODEs) have gained momentum in computer science due to the interest in formal methods for computational biology [35,14,20] and for their capability of accurately approximating large-scale Markovian models [24,37,5,40,30]. This has led to a number of results concerning the important, cross-disciplinary, and longstanding problem of reducing the size of ODE systems (e.g., [32,2,27]) using techniques such as abstract interpretation [18,13] and bisimulation [39,19,26,9,12].…”

Estratégias comutativas para análise de confiabilidade em linha de produtos de software

“…In particular, several model checking techniques have been successfully lifted to work with product lines [Apel et al, 2013b;Chrszon et al, 2016;Classen et al, 2013Classen et al, , 2011Classen et al, , 2014Dubsla et al, 2015;Ghezzi and Molzam Sharioo, 2013;Kowal et al, 2015;Nunes et al, 2012]. In contrast to existing research in this area, our work presents dierent analysis techniques, covering all but one of the strategies identied in the taxonomy by Thüm et al [2014] (the exception being the feature-family-product-based strategy).…”

“…Variability encoding: Previous research has exploited variability encoding (also called conguration lifting) as a technique to produce family-based model checking of product lines [Apel et al, 2011[Apel et al, , 2013bKowal et al, 2015;Post and Sinz, 2008]. von Rhein et al [2016] formalize variability encoding in the context of programming languages, that is, the transformation of compile-time variability into load-time variability.…”

“…Existem diversas abordagens para análise de linha de produtos que adaptam técnicas de análise consagradas de forma a lidar com variabilidade [Thüm et al, 2014]. Em particular, diversas técnicas de model checking foram alçadas à operação em linhas de produtos [Chrszon et al, 2016;Classen et al, 2013Classen et al, , 2011Classen et al, , 2014Dubsla et al, 2015;Ghezzi and Molzam Sharioo, 2013;Kowal et al, 2015;Nunes et al, 2012;Rodrigues et al, 2015;Thüm et al, 2014]. Dentre essas técnicas, o presente trabalho se concentra em análise de conabilidade, que é a vericação de uma propriedade de existência probabilística [Grunske, 2008] e pode ser intuitivamente vista como a probabilidade de que um sistema não falhe.…”