arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.12079cs.SE
收藏

性能反模式:对功耗是天使还是恶魔?

Performance Antipatterns: Angel or Devil for Power Consumption?

Free University of Bozen-Bolzano(博尔扎诺自由大学);Vrije Universiteit Amsterdam(阿姆斯特丹自由大学)

Alessandro Aneggi, Vincenzo Stoico, Andrea Janes

AI总结:

本文研究了性能反模式对能耗的影响,发现部分反模式会显著增加能耗,而另一些则主要延长执行时间而非提高瞬时功率消耗。

摘要:

性能反模式被已知会降低微服务系统响应性,但其对能耗的影响仍鲜有研究。本文实证研究了由Smith和Williams定义的广泛研究的性能反模式是否对能耗产生负面影响。我们实现了十个反模式作为孤立的微服务,并在受控负载条件下评估它们,收集了性能、CPU和DRAM能耗以及资源利用率的同步测量,每种反模式进行30次重复运行。结果表明,尽管所有反模式如预期般降低性能,但只有部分显示出响应时间与增加能耗之间有统计显著的相关性。具体而言,几种反模式达到CPU饱和,无论响应时间上升,功率消耗保持不变;而其他(如不必要的处理、斜坡)则表现出能耗与性能耦合,表明低效。我们的研究结果表明,尽管所有注入的性能反模式如预期般增加响应时间,但只有部分表现出明显的能耗反模式,其中一些情况达到几乎恒定的CPU功率水平,额外的变慢主要转化为更长的执行时间,而非更高的瞬时功率消耗。本研究为识别同时表现出能耗反模式的性能反模式提供了系统基础,并为设计更节能的微服务架构提供了可操作的见解。

英文摘要:

Performance antipatterns are known to degrade the responsiveness of microservice-based systems, but their impact on energy consumption remains largely unexplored. This paper empirically investigates whether widely studied performance antipatterns defined by Smith and Williams also negatively influence power usage. We implement ten antipatterns as isolated microservices and evaluate them under controlled load conditions, collecting synchronized measurements of performance, CPU and DRAM power consumption, and resource utilization across 30 repeated runs per antipattern. The results show that while all antipatterns degrade performance as expected, only a subset exhibit a statistically significant relationship between response time and increased power consumption. Specifically, several antipatterns reach CPU saturation, capping power draw regardless of rising response time, whereas others (\eg Unnecessary Processing, The Ramp) demonstrate energy-performance coupling indicative of inefficiency. Our results show that, while all injected performance antipatterns increase response time as expected, only a subset also behaves as clear energy antipatterns, with several cases reaching a nearly constant CPU power level where additional slowdowns mainly translate into longer execution time rather than higher instantaneous power consumption. The study provides a systematic foundation for identifying performance antipatterns that also behave as energy antipatterns and offers actionable insights for designing more energy-efficient microservices architectures.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑