Abstract: Reverse auction is an effective method for resource allocation of cloud workflow system.But this method causes long execution time and low efficiency because it makes auction separately.By combinatorial reverse auction mechanism,the brother nodes of workflow tasks are combined by combinatorial reverse auction.This brings high auction efficiency.Usually resource price is fixed so that weakly competitive resources providers are hard to sell resources and then results in low resource utilization.To introduce the dynamic pricing mechanism and the dynamic pricing strategy for improving providers competitive ability,a combinatorial reverse auction model was presented by leading the method of dynamic-pricing combinatorial reverse auction into the resources allocation of cloud workflow system,and a dynamic-pricing combinatorial reverse auction algorithm was proposed.Experiments were simulated on Epigenomics workflows,and the results showed that Time Cost (TC) and auction times of combinatorial reverse auction with fixed pricing were lower than reverse auction by 60% and 17% averagely.The combinatorial reverse auction TC of dynamic pricing was lower than fixed pricing by 63% and resource utilization of dynamic pricing was higher than fixed pricing 69% averagely.

Key words: cloud computing, cloud workflow system, resource allocation, dynamic pricing, reverse auction

摘要： 为了引入动态定价机制、动态调整资源价格以提高资源提供商的竞争力,将动态定价组合反向拍卖方法引入云工作流资源分配中,设计了适于云工作流系统的组合反向拍卖模型,提出动态定价组合反向拍卖算法。通过基因序列工作流Epigenomics进行对比实验,结果表明固定价格时组合反向拍卖的时间费用之积和拍卖次数比反向拍卖分别平均降低60%和17%,组合反向拍卖中动态定价的时间费用之积比固定价格平均降低63%,资源利用率平均提高69%。

关键词: 云计算, 云工作流系统, 资源分配, 动态定价, 反向拍卖