Virtual machine consolidation using constraint-based multi-objective optimization

作     者：Terra-Neves, Miguel Lynce, Ines Manquinho, Vasco 

作者机构：Univ Lisbon INESC ID Inst Super Tecn Rua Alves Redol 9 P-1000029 Lisbon Portugal 

出 版 物：《JOURNAL OF HEURISTICS》 (试探法杂志)

年 卷 期：2019年第25卷第3期

页      面：339-375页

核心收录：

学科分类：07[理学] 070104[理学-应用数学] 0701[理学-数学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：national funds through Fundacao para a Ciencia e a Tecnologia (FCT) [UID/CEC/50021/2013, SFRH/BD/111471/2015, CMU/AIR/0022/2017, PTDC/CCIC-OM/31198/2017] Fundação para a Ciência e a Tecnologia [SFRH/BD/111471/2015, CMU/AIR/0022/2017] Funding Source: FCT 

主　　题：Virtual machine consolidation Multi-objective optimization Pseudo-Boolean optimization Live migration Cloud computing 

摘      要：With the blooming of cloud computing, the demand for data centers has been rising greatly in recent years. Their energy consumption and environmental impact has become much more significant due to the continuous growth of data center supply. It is possible to reduce the amount of energy consumed by a data center by shutting down unnecessary servers and maintaining only a subset running, such that it is enough to fulfill the resource demand. With recent advances in virtualization technology, it even became possible to consolidate the workload of multiple under-utilized servers into a single server. However, too aggressive consolidation may lead to significant degradation of data center performance. Therefore, the problem of simultaneously minimizing energy consumption and performance degradation in a data center is a complex and challenging problem. In this paper, a novel multi-objective Boolean optimization encoding for virtual machine consolidation is proposed and several approaches to solve it are described and compared. Moreover, this encoding is extended to consider anti-collocation constraints and the migration of virtual machines that are initially placed. This work is in part motivated by the great improvements in the performance of Boolean optimization solvers, thus increasing their applicability and effectiveness for a wider spectrum of complex problems. In this case, specific techniques are applied to further boost the performance, namely search space reduction by symmetry breaking and heuristic reduction of the instance size. An extensive experimental evaluation shows the suitability of the proposed solution in comparison to the state-of-the-art approaches based on stochastic methods.