Measuring server QoS in open virtual worlds : relating QoS to QoE for OpenSim Servers on the Hyper-GRID
Abstract
To understand the performance of networked computer systems it is necessary to consider client, server and network components. Virtual Worlds are resource intensive interactive systems. As such they are liable to degradations in system performance which have a negative impact upon users’ Quality of Experience (QoE). To efficiently use system resources it is necessary to understand when and how changes in Quality of Service (QoS) impact upon the QoE. This study investigates the factors that affect the performance of Open Virtual World Servers and how these impact upon users. It presents quantitative experimental results that establish factors which affect QoS performance and qualitative results of user QoE. The central research question asks ’to what extent can we identify easily measurable system properties that will enable us to make efficient use of server resources whilst preserving QoE? This paper identifies the Server Frame Time (SFT), the time taken by the server to create a frame, and Server Frames per Second (SFPS), the number of frames created by a server in a second, as key metrics which are easy to measure and relate directly to QoE. Most virtual world servers are allocated resources manually and this allocation is independent of the actual work load. As a result, server resources are under utilised due to variations in load. This approach causes a higher provisioning cost than is necessary. The study in this project establishes a way to determine the usability threshold. Based on this server resources can be tuned to provide appropriate QoE for users while making sure that they are utilised to their full potential.
Citation
Miller , A H D , McCaffery , J P , Allison , C , Dow , L & Oliver , I A 2014 , Measuring server QoS in open virtual worlds : relating QoS to QoE for OpenSim Servers on the Hyper-GRID . in Proceedings of pgNET 2014 . IEEE .
Publication
Proceedings of pgNET 2014
Type
Conference item
Rights
© PGNet 2014.
Collections
Items in the St Andrews Research Repository are protected by copyright, with all rights reserved, unless otherwise indicated.