[1]郑元伟,李子鹏,龙诺亚,等.基于函数关联集的冷启动优化策略研究[J].机械与电子,2025,(11):15-21.
　ZHENG Yuanwei,LI Zipeng,LONG Nuoya,et al.Research on Cold-start Optimization Strategy Based on Function Association Set[J].Machinery & Electronics,2025,(11):15-21.
点击复制

基于函数关联集的冷启动优化策略研究()
分享到：

《机械与电子》[ISSN:1001-2257/CN:52-1052/TH]

卷:
期数:
2025年11期
页码:
15-21
栏目:
研究与设计
出版日期:
2025-11-24

文章信息/Info

Title:
Research on Cold-start Optimization Strategy Based on Function Association Set
文章编号:
1001-2257 ( 2025 ) 11-0015-07
作者:
郑元伟 1 ; 李子鹏 1 ; 2 ; 龙诺亚 1 ; 张 菡 1 ; 张 猛 1 ; 宋 磊 1 ; 王喜宾 3
1. 贵州电网有限责任公司,贵州 贵阳 550002 ;
2. 贵州大学计算机科学与技术学院,贵州 贵阳 550025 ;
3. 贵州理工学院大数据学院,贵州 贵阳 550025
Author(s):
ZHENG Yuanwei1 ;  LI Zipeng1 ;  2 ;  LONG Nuoya1 ;  ZHANG Han1 ;  ZHANG Meng1 ;  SONG Lei1 ;  WANG Xibin3
( 1.Guizhou Power Grid Co. , Ltd. , Guiyang 550002 , China ;?
2.College of Computer Science and Technology , Guizhou University , Guiyang 550025 , China ;?
3.College of Data Science , Guizhou Institute of Technology , Guiyang 550025 , China )
关键词:
Serverless ; 冷启动; 协同预热; 调度优化;  FP-Growth
Keywords:
Serverless ;  cold start ;  collaborative pre-warming ;  scheduling optimization ;  FP-Growth
分类号:
TP393
文献标志码:
A
摘要:
Serverless 计算中的云函数冷启动问题会导致显著的性能开销。现有预热策略在以应用为调度单位时,常因调度粒度过粗而造成资源浪费,且难以对调用模式不清晰的函数进行有效优化。为此,提出一种基于函数关联集的冷启动优化策略 FS-Warm 。该策略引入“函数关联集”作为新的调度粒度,通过挖掘函数间的共现关系(如使用 FP-Growth 算法),将业务逻辑或调用行为上强关联的函数(包括高频函数及其关联的低频或模式不清晰函数)聚合。基于函数关联集进行协同预热与调度,旨在更精准地按需加载资源,并利用已知模式函数的调用行为辅助优化集内其他函数的预热时机,从而有效缓解冷启动问题,提高资源利用率并改善对无清晰调用模式函数的优化效果。实验结果表明,基于函数关联集的冷启动优化策略 FS-Warm 在 Serverless 云函数冷启动率的降低和内存浪费的平衡上均具有较好的表现,在 75 分位点上将冷启动率降至 20.8% ,相较于基线策略优化了 42.4% ,并减少了约 75% 的内存浪费。
Abstract:
The cold start problem of cloud functions in serverless computing can lead to significant performance overhead.Existing prewarming strategies , when using applications as the scheduling unit , often result in resource wastage due to overly coarse scheduling granularity and struggle to effectively optimize functions with unclear invocation patterns.This paper proposes FS-Warm , a cold start optimization strategy based on function association sets.This strategy introduces “ function association sets ” as a new scheduling granularity , mining co-occurrence relationships between functions ( e.g. , using the FP-Growth algorithm ) to aggregate functions that are strongly correlated in business logic or invocation behavior ( including high frequency functions and their associated low frequency or pattern unclear functions ) .By performing collaborative prewarming and scheduling based on function association sets , the strategy aims to load resources more precisely on demand and leverage the invocation behavior of known pattern functions to assist in optimizing the prewarming timing of other functions within the set.This effectively mitigates the cold start problem , improves resource utilization , and enhances optimization for functions without clear invocation patterns.Experimental results show that FS-Warm , the cold start optimization strategy based on function association sets , performs well in reducing the cold start rate of serverless cloud functions and balancing memory wastage.At the 75th percentile , it reduces the cold start rate to 20.8% , a 42.4% improvement over baseline strategies , while cutting memory wastage by approximately 75%.

参考文献/References:

[ 1 ] JANGDA A , PINCKNEY D , BRUN Y , et al.Formal foundations of serverless computing [ J ] .Proceedings of the ACM on programming languages , 2019 , 3 ( OOPSLA ): 149.

[ 2 ] CASTRO P , ISHAKIAN V , MUTHUSAMY V , et al. The rise of serverless computing [ J ] .Communications of the ACM , 2019 , 62 ( 12 ): 44-54.
[ 3 ] YU H F , ROY R B , FONTENOT C , et al.RainbowCake : mitigating cold starts in serverless with layer wise container caching and sharing [ C ] ∥Proceedings of the 29th ACM International Conference on Architectural Support for Programming Languages and Operating Systems , 2024 : 335-350.
[ 4 ] GOLEC M , WALIA G K , KUMAR M , et al.Cold start latency in serverless computing : a systematic review , taxonomy , and future directions [ J ] .ACM Computing surveys , 2025 , 57 ( 3 ): 1-36.
[ 5 ] PAHL C , BROGI A , SOLDANI J , et al.Cloud container technologies : a state-of-the-art review [ J ] .IEEE Transactions on cloud computing , 2019 , 7 ( 3 ):677-692.
[ 6 ] MANNER J , ENDRE M , HECKEL T , et al.Cold start influencing factors in function as a service [ C ] ∥2018 IEEE / ACM International Conference on Utility and Cloud Computing Companion , New York : IEEE , 2018 : 181-188.
[ 7 ] HASSAN H B , BARAKAT S A , SARHAN Q I.Survey on serverless computing [ J ] .Journal of cloud computing , 2021 , 10 ( 1 ): 39.
[ 8 ] SHAFIEI H , KHONSARI A , MOUSAVI P.Serverless computing : a survey of opportunities , challenges , and applications [ J ] .ACM Computing surveys , 2022 , 54 ( 11s ): 1-32.
[ 9 ] SHAHRAD M , FONSECA R , GOIRI I , et al.Serverless in the wild : characterizing and optimizing the serverless workload at a large cloud provider [ C ] ∥Proceedings of the 2020 USENIX Annual Technical Conference , 2020 : 205-218.
[ 10 ] LI Z J , GUO L S , CHENG J G , et al.The serverless computing survey : a technical primer for design archi- tecture [ J ] .ACM Computing surveys ( CSUR ), 2022 ,54 ( 10s ): 1-34.
[ 11 ] HAN J W , PEI J , YIN Y W , et al.Mining frequent patterns without candidate generation : a frequent pattern tree approach [ J ] .Data mining and knowledge discovery , 2004 , 8 ( 1 ): 53-87.

备注/Memo

备注/Memo:
收稿日期: 2025-06-12
基金项目:贵州电网有限责任公司创新项目( GZKJXM2O220069 )
作者简介:郑元伟 ( 1988- ),男,贵州盘州人,硕士,高级工程师,研究方向为电力通信技术、光纤传感技术等;龙诺亚 ( 1991- ),男,贵州贵阳人,研究方向为通信软交换技术、语音平台等;张 菡 ( 1984- ),女,贵州瓮安人,硕士,高级工程师,研究方向为电网信息通信;张 猛 ( 1983- ),男,辽宁鞍山人,研究方向为电力通信网络、电力物联网技术应用等;王喜宾 ( 1985- ),男,河南洛阳人,博士,教授,研究方向为机器学习、服务计算。
更新日期/Last Update: 2025-12-12