The impact of the X Window System upon Ethernet performance

The purpose of this thesis was to investigate the demands placed upon an Ethernet by remote clients of the X Window System. Three methods of investigation were applied - queuing analysis, simulations, and an after-the-fact analysis based upon utilization statistics gathered while monitoring network traffic. Each of these methods was applied to profiles which characterize the primary transactions executed by a user -- the display of textual or graphical information and the input of information either by keyboard or mouse. Accuracy of the data obtained from simulations was confirmed by both the theoretical analysis and the experimental results taken from an actual network. Given the reliability of the output, and considering the flexibility of the simulation tool, simulations were used as the primary instrument for the evaluation. Results from this study reveal that the display of text and the input of graphical information provide insignificant network loads, while the display of graphics and the input of text can produce appreciable loads yet maintain acceptable user response times. In contrast, applications with a high degree of user interaction (i.e., those applications which make extensive use of drag operations) can saturate the network and provide very poor response times with relatively few users. A couple of network configurations are given whereby remote X traffic may be localized to avoid potential network saturation.