A Combinatorial Auction-Based Collaborative Cloud Services Platform  

作　　者：Xiaowei Zhang Bin Li Junwu Zhu 

机构地区：[1]School of Information Engineering, Yangzhou University [2]State Key Laboratory for Novel Software Technology, Nanjing University

出　　处：《Tsinghua Science and Technology》2015年第1期50-61,共12页清华大学学报（自然科学版（英文版）

基　　金：supported by the National Natural Science Foundation of China (Nos. 61070133, 61170201, and 61472344);the Collegiate Natural Science Foundation of Jiangsu Province (Grant No. 11KJD520011);Six talent peaks project in Jiangsu Province (No. 2011-DZXX-032);the Scientific Research Foundation of Graduate School of Jiangsu Province (No. CXZZ13 0901)

摘　　要：In this paper, we present a novel, dynamic collaboration cloud platform in which a Combinatorial Auction（CA）-based market model enables the platform to run effectively. The platform can facilitate expense reduction and improve the scalability of the cloud, which is divided into three layers： The user-layer receives requests from end-users, the auction-layer matches the requests with the cloud services provided by the Cloud Service Provider（CSP）, and the CSP-layer forms a coalition to improve serving ability to satisfy complex requirements of users.In fact, the aim of the coalition formation is to find suitable partners for a particular CSP. However, identifying a suitable combination of partners to form the coalition is an NP-hard problem. Hence, we propose approximation algorithms for the coalition formation. The Breadth Traversal Algorithm（BTA） and Revised Ant Colony Algorithm（RACA） are proposed to form a coalition when bidding for a single cloud service in the auction. The experimental results show that RACA outperforms the BTA in bid price. Other experiments were conducted to evaluate the impact of the communication cost on coalition formation and to assess the impact of iteration times for the optimal bidding price. In addition, the performance of the market model was compared to the existing CA-based model in terms of economic efficiency.In this paper, we present a novel, dynamic collaboration cloud platform in which a Combinatorial Auction（CA）-based market model enables the platform to run effectively. The platform can facilitate expense reduction and improve the scalability of the cloud, which is divided into three layers： The user-layer receives requests from end-users, the auction-layer matches the requests with the cloud services provided by the Cloud Service Provider（CSP）, and the CSP-layer forms a coalition to improve serving ability to satisfy complex requirements of users.In fact, the aim of the coalition formation is to find suitable partners for a particular CSP. However, identifying a suitable combination of partners to form the coalition is an NP-hard problem. Hence, we propose approximation algorithms for the coalition formation. The Breadth Traversal Algorithm（BTA） and Revised Ant Colony Algorithm（RACA） are proposed to form a coalition when bidding for a single cloud service in the auction. The experimental results show that RACA outperforms the BTA in bid price. Other experiments were conducted to evaluate the impact of the communication cost on coalition formation and to assess the impact of iteration times for the optimal bidding price. In addition, the performance of the market model was compared to the existing CA-based model in terms of economic efficiency.

关 键 词：cloud computing coalition formation combinatorial auction ant colony algorithm communication cost 

分 类 号：TP393.09[自动化与计算机技术—计算机应用技术]