Remote desktop systems offers users a practical and convenient way to access interactive graphical working environment of a remote workstation. This thesis first gives an overview of technology, diversitys and mechanisms for remote desktop applications, and focuses on the operation of a remote desktop protocol RDP. To further understand the efficiency of the protocol, an analisys is performed, to determine the impact of traffic generated by user, on underlying network parameters. To be able to determine a user activity in general remote desktop traffic, a reference - common behaviour of a user activity, is first identified in user generated traffic. Reference activities are defined as: no user activity, document editing, browsing and video content viewing task. The traffic load with no user activity, exhibit very low constant bit rate that averages in range of 2 kbit/s to 6 kbit/s, inside wich only data for maintaining the session is transmitted. Document editing traffic shows weak asymmetry and a significant variance of traffic. Maximum bit rate load for that task, is observed between 5 kbit s and 15 kbit/s . With the use of menus the bit rate jumps between 100 kbit/s and 300 kbit/s. Web browser task causes a significant traffic bursts, caused by frequent refreshes of the screen and a very dynamic desktop content. The bit rate is observed from 500 kbit/s up to 1500 kbit/s. Reference task of watching a video content causes low traffic load in the direction to remote desktop server and high traffic loads to client side with a constant bitrate of 30 Mbit/s. Subsequent analysis of transport level traffic flows, is then statistically evaluated for individual reference user activity sample parameters. The results show that for no user activity the average transferred rate up to 5 pak/s. The task of web browsing in average ranges of up to 2.5 Mbit / s and task of watching video up to 7 Mbit/s. By comparing the obtained sample parameters covered and the actual live traffic follows, it is possible to identify user task usage in Remote Desktop connection with 78% certainty of recognition. We find that 92.91% of user sessions is without activity. Editing documents is detected by 6.14% of sessions and 0.89% sessions of online browsing. Usage of video content was not detected.User activity identification is then derived from live real traces of captured remote desktop traffic with a comaparison method of previously determined sample traffic parameters for individual user activity and measured traffic parameters. In the end, using a subjective assessment of quality of service (MOS) delivered an evaluation of the received user experience (QoE) for each reference job depending on network parameters. We find that document editing good user experience, is not disturbed even with a 10% packet loss. Guaranteed transfer bit rate for comfortable work, should be greater than 50 kbit/s. Network delay of 150 ms to 200 ms is considered satisfactory. For satisfactory user experience of web browsing application the network loss of 5% is sustainable. Guaranteed transfer bit rate must be at least 150 kbit / s and the delay should not exceed 90 ms. It is interesting to note that even in ideal settings of network conditions for this task, user experience does not achieve very good results (MOS = 5). For task of watching a video we conclude that even for ideal network condition, user experience of that task never exceeds value of MOS ≈ 3. Here the transfer bit rate should be at least 2 Mbit/s for bearable user experience and network packet loss of more than 2% renders user experience completly unacceptable. From this we can conclude, that RDP protocol is not intended for applications, that show high network load.
|
