Research On End-to-end Transmission Schemes For Screen Updates In Desktop-as-a-Service

In recent years,users’ application demands for some services,such as mobile office and online education,are increasing faster than ever.Under such a situation,an emerging cloud service mode,Desktop-as-a-Service(DaaS),acquires unprecedented attention.DaaS can decouple desktop environments of traditional computers from hardware devices,providing users with a flexible and economical way to access virtual desktops remotely.To maintain the same quality of experience(QoE)as if manipulating local devices,users ought to perceive the changes of contents in virtual desktops reliably and timely.These changes are encapsulated as screen updates and sent by cloud servers to terminal devices over end-to-end connections.Thus,screen updates have the closest relationship with users’ QoE in DaaS.How to improve the end-to-end transmission efficiency of screen updates between cloud servers and terminal devices has become one of the key issues related to users’ QoE in DaaS and determining the application scale and commercial prospect of DaaS.Currently,a majority of desktop access protocols developed by Citrix,Microsoft,Red Hat,et al.,supporting the DaaS business of Amazon,IBM,and so on,still use traditional transmission schemes for the end-to-end delivery of screen updates,which will face the following severe challenges with the development of network technology and the maturity of virtual desktop technology.Firstly,when screen updates are transmitted via networks with different basic architectures and channel characteristics,the transmission efficiency of traditional transmission schemes will be negatively affected by the heterogeneous and dynamic transmission environment.Secondly,advanced DaaS solutions classify screen updates into text,image,and video types,whose differentiated transmission requirements for reliability and timeliness cannot be satisfied by traditional transmission schemes.Besides,they cannot realize dynamic transmission protection for screen updates,either.Finally,traditional transmission schemes are unable to perceive users’ QoE during the delivery of screen updates and they cannot react to the drop of users’ QoE in time,leading to difficulties in catering to users’ requirements for enhancing QoE.To deal with these challenges,this thesis takes the further promotion for the end-to-end transmission performance of screen updates as a means,aiming to guarantee users’ QoE in DaaS.From three bottom-up aspects of end-to-end transmission: transmission environment,transmission requirement,and user experience,three novel transmission schemes are designed for the end-to-end delivery of screens updates.Specific works and innovations are outlined as follows.(1)To better cope with the heterogeneity and dynamicity of the transmission environment,an intelligent and reliable screen updates transmission scheme,dubbed IRTS,based on reinforcement learning and fountain code is designed.It uses the reinforcement learning algorithm,SARSA(State Action Reward State Action),to conduct online learning for the construction of ideal mappings between network states and sending actions,so as to strengthen the capability of bandwidth utilization and the adaptability to the heterogeneous and dynamic transmission environment.To provide a completely reliable transmission service,IRTS employs hybrid automatic repeat request technology to recover lost screen updates,which is promising to shorten the end-to-end delivery delay by reducing the number of unnecessary retransmissions.Experimental results show that IRTS can make reasonable use of bandwidth to achieve higher screen updates transmission efficiency.The changes in network conditions have less impact on the transmission performance of IRTS,indicating that it has a stronger capability to adapt to complex transmission environment and can effectively ensure the transmission quality of screen updates.(2)To meet the differentiated transmission requirements of screen updates with different types for reliability and timeliness,a screen updates transmission scheme adapting encoding strategies and redundancy of fountain code,called AESR,is designed.It evaluates the importance of packets encapsulating different types of screen updates by estimating the loss caused by their drops,and then dynamically determines whether to offer a best-effort transmission service merely or adopt a certain encoding strategy of fountain code according to the obtained loss values and network conditions to protect packets.The main purpose of AESR when adding redundant data is to limit the loss recovery failure probability.If there are too much redundant data to be generated,resulting in affecting the timeliness of screen updates transmission,AESR will appropriately reduce the amount of redundancy.Experimental results show that AESR has advantages in weighing reliability and timeliness during the transmission of screen updates.It can also keep its transmission overhead at a rational level under different network conditions,suggesting that AESR has the potential to better meet the differentiated transmission requirements of screen updates.(3)To ameliorate the situation that traditional transmission schemes cannot perceive users’ QoE in real-time when delivering screen updates,a partially reliable screen updates transmission scheme,named PRTS,based on the perception of users’ QoE is designed.It quantifies users’ QoE by comparing the actual display time and the expected display time of screen updates,and then adjusts the sending rate based on the perception results of users’ QoE and feedback information reflecting network conditions.PRTS just provides a partially reliable transmission service.Only important screen updates will be restored if they are lost,and screen updates that may exceed their display deadline will be discarded proactively.Experimental results show that PRTS can make the actual display time of screen updates closer to the expected display time,indicating that perceiving users’ QoE helps PRTS acquire transmission performance gains.Meanwhile,PRTS can let a terminal device receive more screen updates and have more screen update times,which is conducive to boosting users’ QoE in DaaS.