In a cloud computing environment there are two types of cost associated with the auto-scaling systems: resource cost and Service Level Agreement (SLA) violation cost. The goal of an auto-scaling system is to find a balance between these costs and minimize the total auto-scaling cost. However, the existing auto-scaling systems neglect the cloud client’s cost preferences in minimizing the total auto-scaling cost. This paper presents a cost-driven decision maker which considers the cloud client’s cost preferences and uses the genetic algorithm to configure a rule-based system to minimize the total auto-scaling cost. The proposed cost-driven decision maker together with a prediction suite makes a predictive auto-scaling system which is up to 25% more accurate than the Amazon auto-scaling system. The proposed auto-scaling system is scoped to the business tier of the cloud services. Furthermore, a simulation package is built to simulate the effect of VM boot-up time, Smart Kill, and configuration parameters on the cost factors of a rule-based decision maker.

中文翻译：

---

使用遗传算法在搜索空间中找到针对云预测成本驱动型决策者的最佳解决方案

在云计算环境中，与自动缩放系统相关的成本有两种：资源成本和违反服务水平协议（SLA）的成本。自动缩放系统的目标是在这些成本之间找到平衡，并使总的自动缩放成本最小化。但是，现有的自动缩放系统在最小化总的自动缩放成本时忽略了云客户端的成本偏好。本文介绍了一种成本驱动型决策者，该决策者考虑了云客户端的成本偏好，并使用遗传算法配置了基于规则的系统，以最大程度地减少总的自动扩展成本。拟议的成本驱动型决策者与预测套件一起构成了一种预测性自动缩放系统，其准确度比亚马逊的自动缩放系统高25％。提议的自动扩展系统适用于云服务的业务层。此外，构建了一个仿真程序包来仿真VM启动时间，Smart Kill和配置参数对基于规则的决策者的成本因素的影响。