谢宁1,
张学杰1,
李伟东2,,
1.云南大学信息学院 ??昆明 ??650500
2.云南大学数学与统计学院 ??昆明 ??650500
基金项目:国家自然科学基金(61472345, 61762091, 11663007)，云南省教育厅科学研究基金(2017ZZX228)
详细信息
作者简介:张骥先：男，1980年生，讲师，研究方向为分布式系统、云计算、移动计算
谢宁：女，1991年生，硕士生，研究方向为云计算
张学杰：男，1965 年生，教授，博士生导师，研究方向为高性能计算、可重构计算
李伟东：男，1981年生，副教授，研究方向为组合优化和算法博弈论
通讯作者:李伟东 weidong@ynu.edu.cn
中图分类号:TP302计量
文章访问数:1318
HTML全文浏览量:469
PDF下载量:63
被引次数:0
出版历程
收稿日期:2018-06-13
修回日期:2018-12-24
网络出版日期:2019-01-02
刊出日期:2019-05-01
Supervised Learning Based Truthful Auction Mechanism Design in Cloud Computing
Jixian ZHANG1,Ning XIE1,
Xuejie ZHANG1,
Weidong LI2,,
1. School of Information Science and Engineering, Yunnan University, Kunming 650500, China
2. School of Mathematics and Statistics, Yunnan University, Kunming 650500, China
Funds:The National Natural Science Foundation of China (61472345, 61762091, 11663007), The Scientific Research Foundation of Department of Education of Yunnan Province (2017ZZX228)
摘要
摘要:使用拍卖方式来进行资源分配可以使得资源提供商获得更大的收益，是云计算领域近年来研究的重点之一。但资源分配问题是NP难的，无法在多项式时间内求解，现有研究主要通过近似算法或启发式算法来实现资源分配，但存在算法耗时长，与最优解相比准确度低的缺点。监督学习中分类及回归思想可对多维云资源分配问题进行建模和分析，针对不同问题规模，该文提出基于线性回归、逻辑回归、支持向量机的3种资源分配算法，并且基于临界值理论设计了支付价格算法，从而确保拍卖机制的可信性。在社会福利、分配准确率、算法执行时间、资源利用率等多个方面进行测试分析，取得了很好的效果。
关键词:云计算/
资源分配/
机制设计/
监督学习
Abstract:Auction based resource allocation can make resource provider get more profit, which is a major challenging problem for cloud computing. However, the resource allocation problem is NP-hard and can not be solved in polynomial time. Existing studies mainly use approximate algorithms or heuristic algorithms to implement resource allocation in auction, but these algorithms have the disadvantages of low computational efficiency or low allocate accuracy. In this paper, the classification and regression of supervised learning is used to model and analyze multi-dimensional cloud resource allocation, for the different scale of problem, three resource allocation predict algorithms based on linear regression, logistic regression and Support Vector Machine (SVM) are proposed. Through the learning of the small-scale training set, the predict model can guarantee that the social welfare, allocation accuracy, and resource utilization in the feasible solution are very close to the optimal allocation solution. The payment price algorithm based on the critical value theory is proposed which ensure the truthful property of the auction mechanism design. Final experimental results show that the proposed scheme has good effect for resource allocation in cloud computing.
Key words:Cloud computing/
Resource allocation/
Mechanism design/
Supervised learning
PDF全文下载地址:
https://jeit.ac.cn/article/exportPdf?id=481325fd-cb46-459d-a264-6a04d3d3b63c
