Implementing Design Diversity Using Portfolio Thinking to Dynamically and Adaptively Manage the Allocation of Web Services in the Cloud

Implementing Design Diversity Using Portfolio Thinking to Dynamically and Adaptively Manage the Allocation of Web Services in the Cloud
复制标题

DOI:
10.1109/tcc.2015.2415783
复制
发表时间:
2015-07
影响因子:
6.5
通讯作者:
Faisal Alrebeish;Rami Bahsoon
Faisal Alrebeish;Rami Bahsoon
中科院分区:
计算机科学2区
文献类型:
--
作者:
Faisal Alrebeish;Rami Bahsoon

文献摘要

被引文献

相似文献

我们将云视为一个交易Web服务实例的市场,这些实例可以被Web应用程序“租用”。我们认为,在这个市场上,应用程序可以通过从多个云销售商那里选择Web服务实例来“购买”多样性。通过使Web服务实例的选择和分配多样化,应用程序可以潜在地提高其可靠性并降低与违反服务级别协议(SLA)相关的风险。我们提出了一种新的、动态的、自适应的方法来实现云市场中的设计多样性。该方法使用投资组合理论来构建一个多样化的Web服务实例组合,这些实例来自多个云提供商。通过实例说明了该方法的适用性。受控实验也被用来(I)测试方法在将违反SLA的风险降至最低方面的有效性;(Ii)模拟方法在应对市场状况和风险变化时的动态和适应性行为;(Iii)评估分配决策对风险的敏感性及其与其他候选者的相关性;以及(Iv)评估方法的可扩展性及其在极端情况下对降低风险的影响。
We view the cloud as a marketplace for trading instances of web services, which can be “leased” by web applications. We argue that applications can “buy” diversity by selecting instances of web services from multiple cloud sellers in this market. By diversifying the selection and allocation of web service instances, an application can potentially improve its dependability and reduce risks associated with service level agreement (SLA) violations. We propose a novel, dynamic and adaptive approach for implementing design diversity in the cloud market. The approach uses portfolio theory to construct a diversified portfolio of web service instances, which are traded from multiple cloud providers. We illustrate the applicability of the approach. Controlled experiments are also used to (i) test the approach effectiveness in minimizing the risk of SLA violation; (ii) simulate the dynamic and adaptive behaviour of the approach in responding to changes in the market conditions and risk; (iii) evaluate the sensitivity of the allocation decisions to risk and its correlation with other candidates and (iv) evaluate the scalability of the approach and its ramifications on risk reduction under extreme scenarios.