[Paper Review] Multi-factorial Optimization for Large-scale Virtual Machine Placement in Cloud Computing

Zhengping Liang, Jian Zhang+2 more|arXiv (Cornell University)|Jan 18, 2020

IoT and Edge/Fog Computing52 references4 citations

TL;DR

This paper proposes a multi-factorial evolutionary algorithm (MFEA) with a greedy-based allocation and re-migration/merge operator to solve large-scale virtual machine placement (LVMP) in heterogeneous cloud environments. It significantly reduces optimization time—achieving 37.61s for 250,000 VMs—while maintaining high resource utilization and low deployment costs, outperforming HGAPSO and RGGA in scalability and real-time performance.

ABSTRACT

The placement scheme of virtual machines (VMs) to physical servers (PSs) is crucial to lowering operational cost for cloud providers. Evolutionary algorithms (EAs) have been performed promising-solving on virtual machine placement (VMP) problems in the past. However, as growing demand for cloud services, the existing EAs fail to implement in large-scale virtual machine placement (LVMP) problem due to the high time complexity and poor scalability. Recently, the multi-factorial optimization (MFO) technology has surfaced as a new search paradigm in evolutionary computing. It offers the ability to evolve multiple optimization tasks simultaneously during the evolutionary process. This paper aims to apply the MFO technology to the LVMP problem in heterogeneous environment. Firstly, we formulate a deployment cost based VMP problem in the form of the MFO problem. Then, a multi-factorial evolutionary algorithm (MFEA) embedded with greedy-based allocation operator is developed to address the established MFO problem. After that, a re-migration and merge operator is designed to offer the integrated solution of the LVMP problem from the solutions of MFO problem. To assess the effectiveness of our proposed method, the simulation experiments are carried on large-scale and extra large-scale VMs test data sets. The results show that compared with various heuristic methods, our method could shorten optimization time significantly and offer a competitive placement solution for the LVMP problem in heterogeneous environment.

Motivation & Objective

• To address the scalability and high time complexity limitations of traditional evolutionary algorithms (EAs) in large-scale virtual machine placement (LVMP) problems.
• To apply multi-factorial optimization (MFO) to simultaneously optimize multiple correlated VM placement tasks in heterogeneous cloud environments.
• To reduce deployment costs by formulating a cost-based VMP problem within the MFO framework.
• To enhance solution quality and convergence speed using a greedy-based allocation operator and a re-migration/merge operator for integrated solution generation.
• To evaluate the method on large-scale and extra-large-scale test datasets, demonstrating real-time performance and scalability.

Proposed method

• Formulates the deployment cost-based VMP problem as a multi-factorial optimization (MFO) problem to enable simultaneous optimization of multiple VM placement tasks.
• Develops a multi-factorial evolutionary algorithm (MFEA) enhanced with a greedy-based allocation operator to improve initial solution quality and convergence speed.
• Introduces a re-migration and merge operator to refine and integrate solutions from individual MFO tasks into a unified, high-quality placement configuration.
• Employs a population-based search strategy that leverages shared genetic traits across multiple related optimization tasks to exploit implicit parallelism.
• Uses a heterogeneous physical server (PS) environment with diverse CPU, RAM, and disk configurations to model real-world cloud data centers.
• Employs simulation experiments on large-scale (up to 50,000 VMs) and extra-large-scale (up to 250,000 VMs) datasets to validate performance.

Experimental results

Research questions

• RQ1 Can multi-factorial optimization (MFO) effectively reduce the time complexity and improve scalability of evolutionary algorithms for large-scale virtual machine placement (LVMP)?
• RQ2 How does the proposed MFEA with greedy-based allocation outperform existing heuristic and metaheuristic methods in terms of optimization time and solution quality?
• RQ3 To what extent does the re-migration and merge operator enhance solution integration and overall placement efficiency?
• RQ4 How does the proposed method scale with increasing VM workload sizes, particularly beyond 100,000 VMs?
• RQ5 Can the MFO-based approach maintain high resource utilization and low deployment costs in heterogeneous cloud environments under large-scale workloads?

Key findings

• The proposed method achieved an optimization time of only 37.61 seconds for a test set of 250,000 VMs, significantly outperforming HGAPSO (4319.83s) and RGGA (2323.15s).
• On the extra-large-scale dataset EDS6 (250,000 VMs), the method achieved 85.65% average resource utilization, compared to 73.72% for HGAPSO and 82.20% for RGGA.
• The method reduced deployment cost to 114,808.82 units on EDS6, compared to 154,519.62 for HGAPSO and 125,991.47 for RGGA.
• The average running time increased by only about 7 seconds per 50,000 additional VMs, demonstrating strong scalability and stability.
• The method maintained high solution quality with an average comprehensive utilization rate of 85.65% and a low number of activated physical servers (26,453.77) on EDS6.
• The re-migration and merge operator effectively integrated task-specific solutions into a globally optimal placement configuration, improving solution quality and convergence.