亚马逊竞价型云服务定价策略的分析

On Pricing Strategy Analysis for Amazon Spot Cloud Service

当今,移动互联网行业飞速发展,云服务提供了大量方便易用的云资源.在我国提出"互联网+"的概念以及推行了一系列鼓励创业的政策以后,企业和研究机构对于云服务的需求越来越大.云服务提供商众多,其定价机制和服务种类各不相同,存在着巨大的竞争和可挖掘的市场.亚马逊公司是如今云计算领域中的巨头,其中,竞价型云服务以其易用性和低廉价格受到了广大用户的欢迎.因此,研究其定价模式可以帮助云服务供应商完善其定价方法、获取更多盈利,另一方面,帮助用户选择适合自己的付款模式,节约成本.本次研究以亚马逊竞价型云服务作为对象,将亚马逊官方提供的混杂价格历史做了整理、筛选、可视化以及数据统计.预处理后,输入到KNN分类器和k-means分类算法中,实现了分类的功能,通过两种分类方法进行比对,通过Boosting算法投票选出典型类别.另外,提出了一种补齐不同云服务产生价格时间点的方法,可以辅助提高分类的准确性,以便找出最典型的价格轨迹进行统计分析、建模,提取价格变化的共同特征,更加精确地推测定价机制.

In the rapid development of mobile Internet nowadays,Cloud services provide lots of Cloud resources which are easy to use.In particular,demand for utilizing Cloud services has grown by companies and research institutions,after the concept of"Internet +"in China was brought forward,and a series of policies were implemented to encourage entrepreneurship. Amazon is a competitive provider in the Cloud computing market,and Spot Instances( SIs) of AWS is a competitive option for public Cloud users for a lower price. Therefore,the research on the pricing model can help Cloud service providers to improve their pricing mechanism and obtain more profits,as well as helping Cloud users to choose Instance. This study is built on the Amazon Spot Instance. We mainly clean up the mixed dataset for Amazon Spot Instance,and finish the visualization and data statistics of the spot price. After pre-processing,inputting data to the KNN classifier and k-means clustering algorithm,getting the basic classification result,we compare the two kind of methods and use the Boosting approach to vote for the best answer. In addition,We propose a method to fill the time set,which helps to improve the accuracy of classification in order to find out the most typical price trace. In the future,this study will do statistical analysis,modeling,and extraction of common features of price trade,to speculate pricing mechanism.