ABSTRACT

This paper aims to allocate cloud resources (i.e., Virtual Machines: VM) to customers by using an auction-based multi-agent system. Adding the concept of group-buying to auction makes the possibility of having more winners. Group-buying offers resources at a reduced price on the condition that the number of purchases reaches a minimum number. Subsequently, the rate of successful deals, the speed of reaching agreements and monetary utility of the participants (i.e., providers and customers) will be increased. Due to having a massive number of clients with different characteristics and providers with various VM types in the cloud environment, it is obligatory to make the process of group-buying auction fully automated. By doing this automation, customers will be relieved of serious troubles such as coalition for group-buying, comparing the resources, bidding at auctions and so on. Three main requirements of this automation are as follows: (1) clustering similar customers in rational groups, (2) selecting the most suitable resource for each customer/group of customers, and (3) bidding during auctions on behalf of each customer/group of customers. A particular algorithm for each one of these requirements is presented in this paper. Through simulation experiments, the results of evaluations showed how significantly the proposed cloud market system improved the success rate of customers, the average time taken in resource selection, the average of providers’ financial utility, and the average of customers’ financial utility.

摘要

本文旨在通过使用基于拍卖的多代理系统将云资源（即虚拟机：VM）分配给客户。将团购概念添加到拍卖中可以使有更多的获胜者。在购买数量达到最小数量的情况下，团体购买以优惠的价格提供资源。随后，将提高交易成功率，达成协议的速度以及参与者（即提供者和客户）的货币效用。由于云环境中有大量具有不同特征的客户端和具有各种VM类型的提供程序，因此必须使团购拍卖过程完全自动化。通过执行这种自动化操作，客户将免于严重的麻烦，例如结成团购联盟，比较资源，在拍卖中竞标等等。此自动化的三个主要要求如下：（1）将相似的客户聚集在合理的组中；（2）为每个客户/一组客户选择最合适的资源；（3）在拍卖期间代表每个客户/组进行投标客户。本文针对这些需求中的每一项提出了一种特殊的算法。通过仿真实验，评估结果显示了所提出的云市场系统在多大程度上提高了客户的成功率，资源选择的平均时间，提供商的金融效用的平均值以及客户的金融效用的平均值。（2）为每个客户/一组客户选择最合适的资源，以及（3）在拍卖期间代表每个客户/一组客户进行投标。本文针对这些需求中的每一项提出了一种特殊的算法。通过仿真实验，评估结果显示了所提出的云市场系统在多大程度上提高了客户的成功率，资源选择的平均时间，提供商的金融效用的平均值以及客户的金融效用的平均值。（2）为每个客户/一组客户选择最合适的资源，以及（3）在拍卖期间代表每个客户/一组客户进行投标。本文针对这些需求中的每一项提出了一种特殊的算法。通过仿真实验，评估结果显示了所提出的云市场系统在多大程度上提高了客户的成功率，资源选择的平均时间，提供商的金融效用的平均值以及客户的金融效用的平均值。