Research And Design Of Multimedia Desktop Delivery System Based On Virtualization Technology

Nowadays, with the rapid development of cloud computing and software-defineddata center, some of, even most of desk computers and its distributed computingenvironment for various office uses may be replaced by centralized computingenvironment based on desktop virtualization technology. However, in a relative earlystage of commercialization, desktop virtualization needs to greatly improve degree ofstandardization and adaptability for various applications. For example, the availabilityand cost-effective of heavy load applications in desktop virtualization environment arenot nearly so as that in distributed computing environment based on computer.If this problem can be adequately resolved, or at least can be proposed allsolutions, we can theoretically build a new computing fabric and algorithm in desktopvirtualization environment.At the same time,it will break through the significantbottleneck of applying desktop virtualization, becoming the milestone of promotingthe important branch in cloud computing area to practical application.There are two problems for heavy load applications in desktop virtualizationenvironment: smoothly playing of streaming media applications such as audios andvideos, as well as hacking run of large image processing software.These two napplications have different computation structures and different keyindicators, so they need different solving points and different solutions. Here weprimarily analyze and discuss the latter one, hacking run large image processingsoftware,as well as its technological difficulties and tentative scheme of solution.At present, the mainstream solutions of this problem are VMGL、VMware, Xen,Virtual Box and so on in domestic market. Xen and Virtual Box are open source projects, which can be changed. But their solutions tightly couple with their ownvirtual machines.VMware has excellent solution for3D graphics acceleration, but thissolution is built-in its product, which means if you want to use this solution, you mustinstall the same VMware product in all the host machines, limiting the flexibility ofinterior collocation of rendering cloud.In the international market, GPU virtualization is uaually used to hacking runlarge image processing software in desktop virtualization environment, which is toodifficult and expensive to be popularized. Here we consider using computing structureoptimization, which takes full advantage of all the computing and storage units in thewhole visualization platform by using dynamic resource allocation algorithm, tohacking run large image processing software in desktop virtualization environment.In this paper, we primarily research a multimedia desktop delivery method andsystem based on virtualization infrastructure, composed of virtual device interfacesystem, diskless workstation system and united storage system. Because virtual deviceinterface system and diskless workstation system are both use network storage, wecould unify these two storage modes under virtual device interface background system,implementing the following functions: supporting both desktop delivery of virtualdevice interface and desktop delivery of diskless workstation; delivering the samesystem mirror to diskless terminals with different hardware configuration;automatically adapting terminals with desktop launch under virtual device interfacebackground system. This solution combines advantages of both virtual deviceinterface system and diskless workstation system, implementing the followingfunctions: achieving the requirement of effectively supporting heavy applicationsunder virtual device interface system such as graphic image,3D, multimedia and soon; running users’desktops under either virtual device interface desktop or disklessworkstation desktop as their wishes in the same operating environment;besides,savingthe extra cost used to buy diskless workstation system, because the system supportsnot only virtual device interface system, but also diskless workstation system.