Publication No 40642

Author(s)

Kühn, P.J.; Ezzat Mashaly, M.

Title

Load Balancing in Distributed Cloud Data Center Configurations - Performance and Energy Efficiency

Abstract

In this contribution two cloud server clusters are considered which process virtualized user service requests defined as Virtual Machines (VM) operated under the Hypervisor control. Load Balancing (LB) is applied to avoid temporary overloads and to enforce negotiated Service Level Agreements (SLA) defined by means and percentiles of processing delays. Two novel LB strategies are defined through which the two server clusters perfform job processing cooperatively through mutual job overflows by a "Local Server System First" (LSSF) and through a "Shortest Response Time First" (SRTF) strategy, respectively. The cooperation operation is performed by VM migration at the instant of VM scheduling by the Hypervisor. Both LB models are defined by queuing systems which are analyzed by the method of Markov-Chains. Energy efficiency has been analyzed by the authors through server consolidation, server sleep modes, and through Dynamic Voltage and Frequency Scaling (DVFS), c.f. [1-5]. In this contribution another method is studied which is based on a flexible VM migration to a common server cluster by which the total number of servers can be reduced making use of the effect of the economy of scale by server aggregation.

Year

2017

Reference entry

Kühn, P.J.; Ezzat Mashaly, M.
Load Balancing in Distributed Cloud Data Center Configurations - Performance and Energy Efficiency
Proceedings of the 6th International Workshop on Energy-Efficient Data Centres (E2DC), Hongkong, May 2017, pp. 296-301

BibTex file

Download  [BIBTEX]

Full Text

Download  [PDF]