Incorporating Ceteris Paribus Preferences in Multiobjective Virtual Machine Placement

Abstract

Proper handling of preferences in multiobjective evolutionary algorithms is essential for the algorithms' success in real-life applications. While there has been a tremendous work addressing preferences in evolutionary algorithms, the issue of the exact interpretation of decision maker's (DM) preferences and how it affects the performance of evolutionary algorithms has received little attention. One interpretation of preferences that has received significant attention lately by the AI community and is believed to be exercised naturally by decision makers is the Ceteris Paribus (all else being equal) interpretation. In this paper, we adopt the notion of Ceteris Paribus as an interpretation for the DM preferences and incorporate it in a constrained multiobjective problem known as virtual machine placement (VMP). VMP is an essential multiobjective problem in the design and operation of cloud data centers concerned about placing each virtual machine to a physical machine (a server) in the data center. We propose a variant of the NSGA-II that promotes Ceteris Paribus preferred solutions and evaluate its applicability. Our experiment results show that this variant was able to return preferred solutions at almost no extra cost when compared to NSGA-II.

Keywords

• Virtual machine placement
• MOEA
• preference-based GA
• Ceteris Paribus