Um catálogo de medidas de desempenho para avaliação de sistemas autoadaptativos

Abstract

Self-adaptive systems (SAS) are systems capable of evaluating their behavior at runtime and changing it when deemed necessary to avoid the presence of failures. Many of these systems require high processing power to successfully carry out their dynamic and complex operations, maintaining the level of response to their users. As a way of helping SAS to guarantee its quality, quality assessments can be carried out, which can be both a qualitative and quantitative assessment based on measures of software, which seeks to present the perspectives of the system’s attributes. Because it is an essential feature and is part of the majority of trade-offs present in SAS, performance becomes important in the general assessment of SAS in terms of adaptations and assessing their impacts. Thus, the goal of this work is to build a catalog of performance measures to evaluate self-adaptive systems. A literature review was carried out in this work, identifying the sub-characteristics and performance measures in 32 primary studies selected in this review. Most of the measures identified in the primary studies are related to the sub-characteristics of behavior over time, and some measures related to the use of resources. However, measures related to sub-characteristic capacity have not been identified, which is fundamental in a SAS. It directly impacts the end-user and must be preserved in its adaptations. To validate the built catalog, some measures were implemented in a self-adaptive healthcare system where they were evaluated by data extracted from their reports.