Optimizing multi-VM migration by allocating transfer and compression rate using Geometric Programming

doi:10.1016/j.simpat.2020.102201

Return

Optimizing multi-VM migration by allocating transfer and compression rate using Geometric Programming

2021-01-01

PRE

Garima Singh *

Anil Kumar Singh

DOI：10.1016/j.simpat.2020.102201

Original

Original request for help

Save

Abstract

En 中文

Live VM migration has emerged as a core technology for resource management activities and objectives like fault tolerance, reduced energy consumption, load balancing, and system maintenance. Pre-copy migration strategy is frequently used by most of the existing hypervisors for live VM migration. However, it requires a considerable amount of data transfer. In the case of low network bandwidth availability, the migration process gets adversely affected. Thereby degrading the performance of applications running on migrated VMs. Migrating multiple VMs even requires more data to be transferred, which further faces the challenge of sharing the available network bandwidth. This motivates for designing a better strategy for the live migration of multiple VMs. Moreover, to reduce the amount of data that needs to be transferred, memory compression techniques can be utilized efficiently by providing fast and stable VM memory migration without sacrificing any service quality. This paper presents a new model using Geometric Programming that allocates transfer and compression rates to each VM to minimize the total migration time. We have found that: (a). Memory compression, along with pre-copy migration, improves the performance of the live migration of multiple virtual machines by reducing the total migration time and downtime. (b). The compression rate is dynamically allocated according to the available network bandwidth. So, in an adverse environment, when available bandwidth is less, the memory is compressed using slow compression algorithms to optimize the total migration time. In contrast, when more bandwidth is available, it uses a fast compression algorithm to transfer the data quickly. (c). With the proposed approach, more number of VMs can be migrated in parallel with considerably reduced migration time and downtime when compared with existing strategies.

Keywords:

Virtualization

Virtual Machine

Live Migration

Pre-copy

Optimization

Geometric Programming

AI Summary

Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.