Stream Based Multi-tenant Remote Desktop Virtualization System

Nowadays, there are growing needs of computing resources inenterprises, schools and governments. On the contrary, the resources we canuse are limited. It is efficient to use virtualization technology to multiplexresources. Desktop virtualization trends to be a combination ofvirtualization and remote desktop technology. The existing desktopvirtualization systems suffer performance degradation in terms of responsetime and video quality which do not meet requirement of office andentertainment scenarios. On the other hand, previous remote accessapproaches are designed for standalone architectures and require semanticinformation which is not transparent to OS. So they are not feasible indesktop virtualization systems. In light of this, it is important to research andimplement a high performance and high fidelity remote desktopvirtualization system.In this paper, we propose a streaming based remote interactivityarchitecture for desktop virtualization system. It has theses features:1.Theprototype system resides in the host through intercepting virtual displaydevice, which avoids modifying guest OS. So the system is transparent toguest OS and its applications.2. The system integrates server-pushstreaming mechanism with H.264encoder into virtualization system, whichprovides high quality display with low bandwidth consumption and lowlatency of interaction.3. To support multi-tenant management，the systemadds virtual machine scheduling component.In this paper, the prototype system is better than other remote desktopsystem in tests. Existing benchmarks lack the ability of testing remotedesktop system. This paper also proposed four representing testing scenarios for testing and comparison. The four scenarios are office, webbrowsing, editing photo and video playback. The prototype system got8%CPU overhead and no more than2Mpbs bandwidth under one scenariotesting, and had good scalability under multi scenarios. This papercompared the prototype system with other remote desktop system,including X, VNC, RDP and THINC. The prototype system outperformedother system in video quality. It got94%video quality and had lessbandwidth cost compared with THINC while THINC got close videoquality result with the prototype system. The prototype system alsoachieved lowest response time300ms on average in WAN environmentamong all systems. In management components, we used an offlinealgorithm to find out an optimal reversed resource rate of20%under ourexperimental setup.