Home // International Journal On Advances in Telecommunications, volume 11, numbers 1 and 2, 2018 // View article
Authors:
Ngoc Khan Truong
Christian Pape
Sven Reißmann
Thomas Glotzbach
Sebastian Rieger
Keywords: Cloud Computing; Network Services; Live Migration; Energy Efficiency; Renewable Energy
Abstract:
Virtual infrastructures and cloud services became more and more important over the past years. The abstraction from physical hardware offered by virtualization supports an increased energy efficiency, for example, due to higher utilization of underlying hardware through consolidation. Also, the abstraction enables the ability to geographically move cloud services, e.g., to be able to benefit from lowest available energy prices and renewable energy. This article gives an overview on such migration techniques in distributed private cloud environments. The presented OpenStack-based testbed is used to measure migration costs along with the service quality of virtualized network services. Correspondingly, the article illustrates the impact of high memory and input/output (I/O) load on live migrations of network services and evaluates possible optimization techniques. The results gained from the experiments presented in this article, can be used to evaluate whether network services and virtual resources can be migrated to distant sites to reduce energy costs. A potential benefit of such migrations can be to leverage from fluctuating renewable energies across multiple data center sites. Possible improvements as well as side effects of this use case are presented in the evaluation regarding the live migration of virtual network services. Regarding virtual network services, potential drawbacks can result from additional latency when maintaining and using the virtual services across distant locations. To mitigate these effects, the article describes a way to identify dependencies and affinities between virtual and physical resources based on network flow data. The evaluation used a data set of characteristic networks flows from around one hundred virtual machines of the production environment at Fulda University of Applied Sciences. While respecting these requirements and dependencies, the optimization described in this article used weather data of multiple years of three different distant locations in Germany. Possible improvements of the utilization of renewable energies due adaptive placement and migration of virtual resources were evaluated using this data. Together with the detailed evaluation of the costs of these migrations, which especially rise for I/O-intensive migrations, e.g., for virtual network services, the results of this article can be used to increase the overall energy efficiency of data centers in distributed cloud infrastructures.
Pages: 76 to 86
Copyright: Copyright (c) to authors, 2018. Used with permission.
Publication date: June 30, 2018
Published in: journal
ISSN: 1942-2601