A Survey on Spot Pricing in Cloud Computing

Amazon offers spot instances to cloud customers using an auction-like mechanism. These instances are dynamically priced and offered at a lower price with less guarantee of availability. Observing the popularity of Amazon spot instances among the cloud users, research has intensified on defining the users’ and providers’ behavior in the spot market. This work presents an exhaustive survey of spot pricing in cloud ecosystem. An insight into the Amazon spot instances and its pricing mechanism has been presented for better understanding of the spot ecosystem. Spot pricing and resource provisioning problem, modeled as a market mechanism, is discussed from both computational and economics perspective. A significant amount of important research papers related to price prediction and modeling, spot resource provisioning, bidding strategy designing etc. are summarized and categorized to evaluate the state of the art in the context. All theoretical frameworks, developed for cloud spot market, are illustrated and compared in terms of the techniques and their findings. Finally, research gaps are identified and various economic and computational challenges in cloud spot ecosystem are discussed as a guide to the future research.     

SipaaS: Spot instance pricing as a Service framework and its implementation in OpenStack

Designing dynamic pricing mechanisms that efficiently price resources in line with a provider's profit maximization goal is a key challenge in cloud computing environments. Despite the large volume of research published on this topic, there is no publicly available software system implementing dynamic pricing for Infrastructure as a Service cloud spot markets. This paper presents the implementation of a framework called Spot instance pricing as a Service (SipaaS) that supports an auction mechanism to price and allocate virtual machine instances. SipaaS is an open‐source project offering a set of web services to price and sell virtual machine instances in a spot market resembling the Amazon EC2 spot instances. Cloud providers, who aim at utilizing SipaaS, should install add‐ons in their existing platform to make use of the framework. As an instance, we provide an extension to the Horizon – the OpenStack dashboard project – to employ SipaaS web services and to add a spot market environment to OpenStack. To validate and evaluate the system, we conducted an experimental study with a group of 10 users utilizing the provided spot market in a real environment. Results show that the system performs reliably in a practical test environment. Copyright © 2015 John Wiley & Sons, Ltd.     

LBAS: An effective pricing mechanism towards video migration in cloud-assisted VoD system

Migrating video instances to cloud to form a cloud-assisted VoD system is an efficient way to solve the problem of increasing costs in the traditional VoD system. In this paper, we focus on further decreasing the costs in this system. We propose a pricing mechanism named LBAS to allocate video instances from VoD provider’s local servers to cloud platforms. First, a pricing algorithm which conforms to the principle of market is presented to determine the price of each video instance. Then, an auction algorithm which is based on low bid and replaceable instances is proposed to determine the allocation and payment of video instances. At last, in order to avoid the collusion among bidders in low-bid auction, the AEM algorithm is improved. The simulation results demonstrate that the proposed mechanism makes the VoD provider decrease 10% of the costs compared with the VMC strategy.     

云中截止时间动态分配的工作流调度成本优化算法

现如今,如何在满足截止时间约束的前提下降低工作流的执行成本,是云中工作流调度的主要问题之一。三步列表调度算法可以有效解决这一问题。但该算法在截止时间分配阶段只能形成静态的子截止时间。为方便用户部署工作流任务,云服务商为用户提供了的三种实例类型,其中竞价实例具有非常大的价格优势。为解决上述问题,提出了截止时间动态分配的工作流调度成本优化算法(S-DTDA)。该算法利用粒子群算法对截止时间进行动态分配,弥补了三步列表调度算法的缺陷。在虚拟机选择阶段,该算法在候选资源中增加了竞价实例,大大降低了执行成本。实验结果表明,相较于其他经典算法,该算法在实验成功率和执行成本上具有明显优势。综上所述,S-DTDA算法可以有效解决工作流调度中截止时间约束的成本优化问题。     

云环境下基于组合双向拍卖的动态资源定价

云环境下的市场交易机制缺乏灵活性,且在某些情况下定价不合理。为此,提出一种基于组合双向拍卖的动态资源定价模型,给出云资源分配与定价算法,用户通过响应时间出价,资源提供商根据负载情况要价。仿真实验结果表明,该算法与固定比例的定价算法相比,能提高18%的用户利益与9%的资源提供商利益。     

多实例云计算资源市场下超额预订决策方法

针对现有云供应商数据中心负载率低、云用户需求不确定及多样性的问题,为提高云供应商平均利润,建立了不确定需求下的多实例类型云服务超额预订模型。该模型结合实际云计算资源市场下超额预订对于云供应商负载均衡及云服务等级协议(SLA)的影响,给出超额预订的多重约束条件,提出了各实例类型数量最优分配策略。实验结果表明,采用该模型在预约未使用概率为0.25时,云供应商利润较高,数据中心负载率达到78%,最终确定了各实例类型的最优分配数量。     

云环境下基于神经网络和群搜索优化的资源分配机制

在云环境下,各种闲置资源可以通过池化形成资源池,进而利用虚拟化技术将资源池中的不同资源组合以服务的形式提供给用户使用,因此需要合理而有效的机制来分配资源.针对云环境下资源的特点,将经济学和智能方法相结合,提出了一种基于双向组合拍卖的智能资源分配机制.在该机制中,提出了基于体验质量（quality ofexperience,简称QoE）的威望系统,引入威望衰减系数和用户信誉度,降低拍卖中恶意行为造成的影响,为资源交易提供QoE 支持.对拍卖中的竞价决策,综合考虑多种因素,提出了基于BP 神经网络的竞标价格决策机制,不仅可以合理确定竞标价,而且使价格可以动态适应市场变化.最后,由于组合拍卖胜标确定问题是NP 完全的,因此引入群搜索优化算法,以市场盈余和总体威望为优化目标,得到资源分配方案.仿真研究结果表明,该机制是可行和有效的.     

满足工作流执行时限的可抢占虚拟机实例配置和调度方法研究

随着云计算的迅速发展,将工作流部署到云计算平台已经成为了常见的选择。相比于传统的本地工作流,云工作流不仅要考虑计算时长等要求,还要考虑其产生的经济开销。而云计算服务商为了提高资源利用率,提供了可抢占虚拟机实例这种非常廉价但是不稳定的资源。针对工作流在云计算中的调度和执行问题,提出一种满足工作流执行时限的可抢占虚拟机实例配置和调度方法。该方法使用马尔科夫模型和动态规划方法,对可抢占虚拟机实例的价格进行预测,并得到成本最低的出价策略。同时,结合工作流的执行时限要求,在估计的出价策略下对工作流中使用的实例进行配置。实验结果显示,相比于全部使用按需付费虚拟机实例,该方法在满足工作流执行时限的前提下最高可以节省89.9%的计算成本。     

基于拍卖的虚拟机动态供应和分配算法

当前云计算供应商通过定价算法或类似拍卖的算法来分配他们的虚拟机（VM）实例。然而,这些算法大多要求虚拟机静态供应,无法准确预测用户需求,导致资源未得到充分利用。为此,提出了一种基于组合拍卖的虚拟机动态供应和分配算法,在做出虚拟机供应决策时考虑用户对虚拟机的需求。该算法将可用的计算资源看成是“流体”资源,且这些资源根据用户请求可分为不同数量、不同类型的虚拟机实例。然后可根据用户的估价决定分配策略,直到所有资源分配完毕。基于Parallel Workload Archive（并行工作负载存档）的真实工作负载数据进行了仿真实验,结果表明该方法可保证为云供应商带来更高收入,提高资源利用率。     

基于竞价的租赁车辆资源分配和定价机制

针对目前线上汽车租赁平台使用固定价格的方式来租赁车辆给用户,存在着租赁车辆分配不合理、定价不能及时反映真实的市场供需情况、整体社会福利较低等问题,提出一种基于竞价的租赁车辆分配与定价机制。首先,研究了线上汽车租赁问题模型,抽象出数学模型与社会福利最大目标函数;然后,在租赁车辆分配算法中采用基于最小费用最大流算法的最优化车辆资源分配算法;最后,在价格支付算法中采用了可信的VCG价格算法,计算出用户最终的支付价格。对比传统的先来先服务算法,基于竞价的租赁车辆分配与定价机制在订单成功率方面提升了20%~30%,在平台收益方面增加了30%左右。实验结果表明,所提出的机制具有车辆分配最优化、弹性价格的优势。     

基于博弈竞价的区域综合能源系统运营弹性增强策略

能源系统故障可能导致巨大的经济损失, 因此, 提升能源系统的运营弹性具有重要意义. 为了提高区域综合能源系统在市场机制下的弹性, 本文提出了一种基于博弈的能源竞标策略. 首先, 建立了区域综合能源系统模型,其中包括能源控制中心、能源枢纽、储能系统和用户聚合商. 为了在市场环境下通过价格引导弹性增强, 构建了基于Stackelberg博弈的能源竞标框架, 通过调整能源零售价格和需求响应激励价格等策略, 以减小能源缺口并降低价格波动. 考虑到参与者间的隐私保护, 设计了双层迭代定价算法. 最后, 通过仿真算例说明了所提策略在保证供需平衡的同时有效抑制价格波动, 从而降低经济损失.     

认知网络中基于竞价模型的频谱分配研究*

研究了动态频谱分配的现状,提出基于竞价模型的频谱分配算法设计应关注的三个主要问题。针对目前基于竞价的分配模型中没有保证授权用户的QoS的缺点,通过频谱质量分级和服务质量分级,设计了一个竞价频谱分配方案。在该方案中提出干扰价格的概念,并且把干扰价格作为竞价人竞标价格的重要组成部分。仿真结果表明,干扰价格能保证主用户的QoS。     

A game-theoretic approach to the financial benefits of infrastructure-as-a-service

Financial benefits are an important factor when cloud infrastructure is considered to meet processing demand. The dynamics of on-demand pricing and service usage are investigated in a two-stage game model for a monopoly Infrastructure-as-a-Service (IaaS) market. The possibility of hybrid clouds (public clouds plus own infrastructure) turns out to be essential in order that not only the provider but also the clients have significant benefits from on-demand services. Even if the client meets all demand in the public cloud, the threat of building a hybrid cloud keeps the instance price low. This is not the case when reserved instances are offered as well. Parameters like load profiles and economies of scale have a huge effect on likely future pricing and on a cost-optimal split-up of client demand between either a client’s own data center and a public cloud service or between reserved and on-demand cloud instances.     

基于随机森林和投票机制的大数据样例选择算法

针对大数据样例选择问题,提出了一种基于随机森林（RF）和投票机制的大数据样例选择算法。首先,将大数据集划分成两个子集,要求第一个子集是大型的,第二个子集是中小型的。然后,将第一个大型子集划分成q个规模较小的子集,并将这些子集部署到q个云计算节点,并将第二个中小型子集广播到q个云计算节点。接下来,在各个节点用本地数据子集训练随机森林,并用随机森林从第二个中小型子集中选择样例,之后合并在各个节点选择的样例以得到这一次所选样例的子集。重复上述过程p次,得到p个样例子集。最后,用这p个子集进行投票,得到最终选择的样例子集。在Hadoop和Spark两种大数据平台上实现了提出的算法,比较了两种大数据平台的实现机制。此外,在6个大数据集上将所提算法与压缩最近邻（CNN）算法和约简最近邻（RNN）算法进行了比较,实验结果显示数据集的规模越大时,与这两个算法相比,提出的算法测试精度更高且时间消耗更短。证明了提出的算法在大数据处理上具有良好的泛化能力和较高的运行效率,可以有效地解决大数据的样例选择问题。     

奇偶直方图负载均衡超立方对等云MapReduce模型*

针对传统MapReduce算法结构在处理大数据时,负载均衡性能不理想的缺点,设计了一种具有负载均衡机制的层次MapReduce模型。该模型利用超立方拓扑结构对MapReduce的映射操作进行改进,通过特定算法将八个结构化的数据中心链接到一个对等的云环境结构中,并使用奇偶直方图组合采样方式的均衡划分方法,实现在用户请求下的节点工作负荷指数均衡。最后,基于Hadoop框架对所提算法进行仿真实验,结果显示本文所提算法结构相对于原始MapReduce结构,具有更高的并行计算的资源利用率,以及更佳的容错和负载均衡性能,综合性能得到有效提升。     

Effects of Reducing VMs Management Times on Elastic Applications

Cloud infrastructures provide computing resources to applications in the form of Virtual Machines (VMs). Many applications deployed in cloud resources have an elastic behavior, that is, they change the number of servers (VMs) dynamically, adapting the application to the workload. Scaling-out and scaling-in operations are managed by an auto-scaler module, which can be reactive (adapting the number of VMs to the current workload) or proactive (adapting to the expected future workload). The cloud infrastructure provides a management interface to create (deploy) and destroy (shutdown) server instances, operations that require some time to complete. In this work we evaluate to what extent the reduction of the time required by VM management operations, namely deployment and shutdown, impacts the performance of applications and the behavior of reactive and proactive auto-scaling policies. After establishing several ideal boundaries on the use of resources, we carry out a set of experiments that show how short management times drastically reduce the use of resources, while allowing the application to operate within the required performance bounds.     

基于概率模型检验的云渲染任务调度定量验证

云渲染技术已被广泛应用于影视和动漫等行业.与传统的渲染农场和租赁市场模式不同，云渲染系统依托云计算基础设施提供多种软件服务进行渲染作业的方式，正逐渐成为新兴的计算模式.由于任务执行和资源操作等作业调度对于用户而言是透明的，这要求云渲染系统应具备智能化以实现计算资源优化调度和多端任务管理，并对系统可靠性提出了更高要求.针对这一问题，提出了采用概率模型检验对云渲染系统任务调度进行定量评估.首先，考虑渲染服务失效等因素引发的随机系统异常和指令错误，如文件损坏和渲染任务超时等，提出了基于离散马尔可夫链（DTMC）的概率模型对云渲染系统的文件准备模块、资源请求模块、渲染任务执行模块进行形式化建模；其次，从服务质量属性角度提出了9类验证性质用于定义云渲染系统的可靠性，采用概率计算树逻辑（PCTL）描述检验性质公式并执行工具PRISM计算和验证渲染系统可靠性；最后，结合案例和实验证明了该方法的可行性和有效性，尤其是对改进前后云渲染系统进行定量检验，可用于指导如何进行失效恢复和任务切换.因此，该方法在一定程度上可提高云渲染系统的可靠性.     

混合云市场的计算资源交易模型

基于计算资源共享模型的研究,提出了混合云计算资源市场HCRM（Hybrid Cloud Resource Market）的功能架构,研究了市场管理层的交易管理机制,给出了计算资源的供需质量模型,设计了一种质量感知的双向拍卖算法QaDA（Quality-aware Double Auction）。仿真运行结果表明,与普通的连续双向拍卖算法CDA（Continuous Double Auction）相比,QaDA不仅可以引导用户合理定价,也能获得较高的匹配成功率和较高的交易总额。     

Probabilistic Automated Bidding in Multiple Auctions

This paper presents an approach to develop bidding agents that participate in multiple auctions with the goal of obtaining an item with a given probability. The approach consists of a prediction method and a planning algorithm. The prediction method exploits the history of past auctions to compute probability functions capturing the belief that a bid of a given price may win a given auction. The planning algorithm computes a price and a set of compatible auctions, such that by sequentially bidding this price in each of the auctions, the agent can obtain the item with the desired probability. Experiments show that the approach increases the payoff of their users and the welfare of the market.