[论文解读] Continuous Double Auction Mechanism and Bidding Strategies in Cloud Computing Markets

Xuelin Shi, Ke Xu+2 more|arXiv (Cornell University)|Jul 23, 2013

Cloud Computing and Resource Management参考文献 18被引用 15

一句话总结

本文提出了一种用于云计算市场的连续双重拍卖（CDA）机制，以提高资源分配效率，引入了一种新颖的两阶段出价策略（BH-策略），通过在三种情景下的仿真评估，该策略在剩余收益生成、交易成功率和市场效率方面均优于现有策略。

ABSTRACT

Cloud computing has been an emerging model which aims at allowing customers to utilize computing resources hosted by Cloud Service Providers (CSPs). More and more consumers rely on CSPs to supply computing and storage service on the one hand, and CSPs try to attract consumers on favorable terms on the other. In such competitive cloud computing markets, pricing policies are critical to market efficiency. While CSPs often publish their prices and charge users according to the amount of resources they consume, auction mechanism is rarely applied. In fact a feasible auction mechanism is the most effective method for allocation of resources, especially double auction is more efficient and flexible for it enables buyers and sellers to enter bids and offers simultaneously. In this paper we bring up an electronic auction platform for cloud, and a cloud Continuous Double Auction (CDA) mechanism is formulated to match orders and facilitate trading based on the platform. Some evaluating criteria are defined to analyze the efficiency of markets and strategies. Furthermore, the selection of bidding strategies for the auction plays a very important role for each player to maximize its own profit, so we developed a novel bidding strategy for cloud CDA, BH-strategy, which is a two-stage game bidding strategy. At last we designed three simulation scenarios to compare the performance of our strategy with other dominating bidding strategies and proved that BH-strategy has better performance on surpluses, successful transactions and market efficiency. In addition, we discussed that our cloud CDA mechanism is feasible for cloud computing resource allocation.

研究动机与目标

• 为解决云计算资源分配中的低效问题，提出一种电子拍卖平台，实现动态定价与交易。
• 设计一种连续双重拍卖机制，使买方与卖方能够同时出价，提升市场灵活性与效率。
• 开发一种战略出价框架，以最大化个体利润，同时提升整体市场表现。
• 通过基于仿真的基准测试，评估所提策略相对于现有策略的性能表现。
• 验证CDA机制在真实世界云计算市场情景下的可行性与有效性。

提出的方法

• 制定一种云计算连续双重拍卖（CDA）机制，根据价格与数量实时匹配买入与卖出订单。
• 提出BH-策略，一种两阶段博弈论出价策略，可适应市场状况并优化出价位置。
• 采用包含三种不同市场情景的仿真环境，将BH-策略与主流现有策略进行对比。
• 定义关键性能指标，包括剩余收益、成功交易率与市场效率，以评估机制与策略的性能表现。
• 运用博弈论原理，对动态竞争市场环境中买方与卖方之间的战略互动进行建模。
• 采用多轮拍卖机制，持续更新出价，并根据价格-时间优先规则进行匹配。

实验结果

研究问题

• RQ1 与静态定价模型相比，连续双重拍卖机制在云计算资源分配中如何提升效率与公平性？
• RQ2 在具有多个买方与卖方的动态云计算市场中，有效的出价策略的关键特征是什么？
• RQ3 所提出的BH-策略在剩余收益生成与交易成功率方面相较于现有出价策略表现如何？
• RQ4 CDA机制在多样化云计算市场条件下，能在多大程度上提升市场效率与稳定性？
• RQ5 所提出的拍卖机制是否可在真实世界的云计算服务环境中实现实际部署并实现可扩展性？

主要发现

• BH-策略在生成总市场剩余收益方面显著优于其他出价策略，效率指标有明显提升。
• 在所有三种仿真情景中，BH-策略下的成功交易数量始终高于其他竞争策略。
• CDA机制在真实世界云计算市场中展现出良好的可行性与可扩展性，支持动态与持续交易。
• 当采用BH-策略时，市场效率（以总剩余与最大可能剩余之比衡量）显著提升。
• BH-策略的两阶段特性使其能更好地适应市场波动，从而提升个体与整体福利。
• 仿真结果证实，所提出的CDA机制实现了更高的配置效率，并能更快收敛至均衡价格。