Design and theoretical analysis of virtual machine placement algorithm based on peak workload characteristics

Abstract

Virtual machine (VM) placement is a fundamental problem about resource scheduling in cloud computing; however, the design and implementation of an efficient VM placement algorithm are very challenging. To better multiplex and share physical hosts in the cloud data centers, this paper presents a VM placement algorithm based on the peak workload characteristics, which models the workload characteristics of VMs with mathematical method, and measures the similarity of VMs' workload with VM peak similarity. Avoiding virtual machines whose workload has high correlation are placed together, it places the virtual machines with peak workload staggering at different time together, which achieves better VM consolidation through VM peak similarity. This paper focuses on the mathematical analysis of VM peak similarity, and proves that compared to cosinesimilarity method and correlation-coefficient method, peaksimilarity method is better theoretically. Finally, numerical simulations and algorithm experiments show that our proposed peak-similarity-based placement algorithm outperforms the random placement algorithm and correlationcoefficient-based placement algorithm.

An intelligent power consumption model for virtual machines under CPU-intensive workload in cloud environment

“…Linear power model is widely used in researches including Ye et al (2012), Basmadjian et al (2011), Kansal et al (2010 and Lee and Zomaya (2012). In addition to that, linear model is also popular in evaluating resource scheduling algorithms (Sampaio and Barbosa 2014;Beloglazov et al 2012) and virtual machine placement strategies (Lin et al 2015a). However, Hsu and Poole (2011) surveyed the power data of 177 different types of servers provided by SPECpower_ssj2008 between 2007 and 2010.…”

Energy-performance optimisation for the dynamic consolidation of virtual machines in cloud computing

“…Horri et al (2014) put forward a VM allocation policy based on the CPU utilisation and the minimum correlation with the migrating VM. Lin et al (2017) proposed a method based on the peak load, and formulated a similarity measurement model based on the current load of a VM and its peak load.…”

A cloud server energy consumption measurement system for heterogeneous cloud environments

“…Energy model-based method is a mainstream method to calculate the energy consumption of cloud computing because of its high flexibility and fine granularity [27]. Currently, energy consumption measuring tools for heterogeneous cloud environments [24,25] are still uncommon, with most of the tools only focused on cluster resource utilization and network monitoring, like Ganglia [6] and Nagios [8]. The main research direction is using power model to estimate power consumption, and modeling CPU, memory and disk as three parts.…”