Best performance and reliability for your time: budget-aware search-based optimization of software model refactoring
CoRR(2023)
摘要
Context: Software model optimization is a process that automatically
generates design alternatives, typically to enhance quantifiable non-functional
properties of software systems, such as performance and reliability.
Multi-objective evolutionary algorithms have shown to be effective in this
context for assisting the designer in identifying trade-offs between the
desired non-functional properties. Objective: In this work, we investigate the
effects of imposing a time budget to limit the search for design alternatives,
which inevitably affects the quality of the resulting alternatives. Method: The
effects of time budgets are analyzed by investigating both the quality of the
generated design alternatives and their structural features when varying the
budget and the genetic algorithm (NSGA-II, PESA2, SPEA2). This is achieved by
employing multi-objective quality indicators and a tree-based representation of
the search space. Results: The study reveals that the time budget significantly
affects the quality of Pareto fronts, especially for performance and
reliability. NSGA-II is the fastest algorithm, while PESA2 generates the
highest-quality solutions. The imposition of a time budget results in
structurally distinct models compared to those obtained without a budget,
indicating that the search process is influenced by both the budget and
algorithm selection. Conclusions: In software model optimization, imposing a
time budget can be effective in saving optimization time, but designers should
carefully consider the trade-off between time and solution quality in the
Pareto front, along with the structural characteristics of the generated
models. By making informed choices about the specific genetic algorithm,
designers can achieve different trade-offs.
更多查看译文
关键词
optimization,software,budget-aware,search-based
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要