Performance modelling of virtualized servers

Gemikonakli O., EVER E., Gemikonakli E.

12th UKSim International Conference on Modelling and Simulation, UKSim 2010, Cambridge, England, 24 - 26 March 2010, pp.434-438, (Full Text) identifier

  • Publication Type: Conference Paper / Full Text
  • Doi Number: 10.1109/uksim.2010.86
  • City: Cambridge
  • Country: England
  • Page Numbers: pp.434-438
  • Keywords: Analytical modelling, Markov process, Performance evaluation, Queuing theory, Server virtualization
  • Middle East Technical University Northern Cyprus Campus Affiliated: Yes

Abstract

The increase on the demand of IT resources diverted enterprises to the use of costly physical servers. However, besides cooling and power cost, most of the times, physical servers are poorly utilized. Server virtualization is a way of better utilizing such resources and reducing power consumption. In this paper, a performance model for two virtual servers implemented on a single physical machine is presented. The proposed model is analyzed using two-dimensional Quasi Birth and Death Processes. An iterative algorithm is used in computing the steady state probabilities of the queuing system considered. The state probabilities are then used in evaluating performance measures of the virtualized servers; mean queue length, response time and utilization. © 2010 IEEE.