An Intelligent Straggler Traffic Management Framework for Sustainable Cloud Environments

IF 3 3区 计算机科学 Q2 COMPUTER SCIENCE, HARDWARE & ARCHITECTURE
Smruti Rekha Swain;Deepika Saxena;Jatinder Kumar;Ashutosh Kumar Singh;Chung-Nan Lee
{"title":"An Intelligent Straggler Traffic Management Framework for Sustainable Cloud Environments","authors":"Smruti Rekha Swain;Deepika Saxena;Jatinder Kumar;Ashutosh Kumar Singh;Chung-Nan Lee","doi":"10.1109/TSUSC.2024.3393357","DOIUrl":null,"url":null,"abstract":"Large-scale computing systems in the modern era distribute tasks into smaller units that can be executed simultaneously to speed up job completion and decrease energy usage. However, cloud computing systems encounter a significant challenge called the Long Tail problem, where a small subset of slow-performing tasks hinders the overall progress of parallel job execution. This behavior leads to longer service response times and reduced system efficiency. This paper introduces a novel approach called Stochastic Gradient Descent with Momentum-driven Neural Network to analyze and classify heterogeneous tasks as either stragglers or non-stragglers. The straggler tasks are further categorized into Resource Hunter and Long-Tail stragglers based on their specific resource requirements. A traffic management policy is implemented to schedule and assign resources among user job requests, considering the task category, to achieve parallelism and improve sustainability within the cloud infrastructure. Extensive simulations are conducted using the Google Cluster Dataset (GCD) to assess the effectiveness of the proposed framework. The results obtained from these simulations are then compared to state-of-the-art techniques. The experimental findings demonstrate significant reductions in power consumption, carbon emissions, active servers, conflicting servers, and VM migration up to 55.16%, 49.76%, 35%, 25.7%, and 87.29%, respectively. Moreover, there has been an enhancement in resource utilization by up to 78.31%, accompanied by a decrease in execution time of up to 67.74%.","PeriodicalId":13268,"journal":{"name":"IEEE Transactions on Sustainable Computing","volume":"10 1","pages":"82-94"},"PeriodicalIF":3.0000,"publicationDate":"2024-04-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"IEEE Transactions on Sustainable Computing","FirstCategoryId":"94","ListUrlMain":"https://ieeexplore.ieee.org/document/10508125/","RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"COMPUTER SCIENCE, HARDWARE & ARCHITECTURE","Score":null,"Total":0}
引用次数: 0

Abstract

Large-scale computing systems in the modern era distribute tasks into smaller units that can be executed simultaneously to speed up job completion and decrease energy usage. However, cloud computing systems encounter a significant challenge called the Long Tail problem, where a small subset of slow-performing tasks hinders the overall progress of parallel job execution. This behavior leads to longer service response times and reduced system efficiency. This paper introduces a novel approach called Stochastic Gradient Descent with Momentum-driven Neural Network to analyze and classify heterogeneous tasks as either stragglers or non-stragglers. The straggler tasks are further categorized into Resource Hunter and Long-Tail stragglers based on their specific resource requirements. A traffic management policy is implemented to schedule and assign resources among user job requests, considering the task category, to achieve parallelism and improve sustainability within the cloud infrastructure. Extensive simulations are conducted using the Google Cluster Dataset (GCD) to assess the effectiveness of the proposed framework. The results obtained from these simulations are then compared to state-of-the-art techniques. The experimental findings demonstrate significant reductions in power consumption, carbon emissions, active servers, conflicting servers, and VM migration up to 55.16%, 49.76%, 35%, 25.7%, and 87.29%, respectively. Moreover, there has been an enhancement in resource utilization by up to 78.31%, accompanied by a decrease in execution time of up to 67.74%.
面向可持续云环境的智能滞留者流量管理框架
现代的大规模计算系统将任务分配到更小的单元中,这些单元可以同时执行,以加快任务完成速度并减少能源消耗。然而,云计算系统遇到了一个重要的挑战,称为长尾问题,其中一小部分执行缓慢的任务阻碍了并行作业执行的整体进度。这种行为会导致服务响应时间变长,降低系统效率。本文介绍了一种基于动量驱动神经网络的随机梯度下降方法来对异构任务进行离散和非离散的分析和分类。掉队任务根据其特定的资源需求进一步分为资源猎手任务和长尾掉队任务。在考虑任务类别的情况下,实现流量管理策略来在用户作业请求之间调度和分配资源,以实现并行性并提高云基础架构内的可持续性。利用谷歌聚类数据集(GCD)进行了大量模拟,以评估所提出框架的有效性。然后将从这些模拟中获得的结果与最先进的技术进行比较。实验结果表明,功耗、碳排放、活动服务器、冲突服务器和VM迁移的显著降低分别达到55.16%、49.76%、35%、25.7%和87.29%。此外,资源利用率提高了78.31%,同时执行时间减少了67.74%。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
求助全文
约1分钟内获得全文 求助全文
来源期刊
IEEE Transactions on Sustainable Computing
IEEE Transactions on Sustainable Computing Mathematics-Control and Optimization
CiteScore
7.70
自引率
2.60%
发文量
54
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信