Virtualization and the client conundrum

If you view system virtualization as an escalating priority, this is a good time to stop and think about client systems and applications before dipping your brush and drawing that first block in your grand virtualization architecture.

Free IT resource Open Source Business Conference (OSBC) May 22-23, 2007 Sponsored by OSBC Free IT resource TechNet: More ways to know it, share it, and keep it running. Sponsored by Microsoft

My visions for VOA (virtualization oriented architecture) hold that operating system instances are to be regarded as services, which doesn’t really address the client. Right now, you probably operate a network of traditional fat PC or Mac clients (equipped to run an OS and applications locally), which have back-end servers mostly doing the old file/print/collaboration thing. Unless you run a very small shop, that traditional client/server model will siphon away some of the bottom-line benefits associated with VOA.

So, if you don’t have your client strategy mapped out in advance, you’ll have no way to size, meter, and tune your server virtualization solutions.

One tried-and-true approach is to think thin. Thin-client systems push the bulk of the user-generated workload, or all of it, to servers. Each thin client interacts with a back-end process that virtualizes anything from a user session in a single instance of a multi-user OS, such as Microsoft Terminal Server, Citrix, or Unix X Window System, to a complete virtual machine with its own OS instance, such as VMware ESX Server. Thin-client technology may not seem glamorous, but it is woefully underutilized. Whatever you decide to do on the server side of your VOA plan, thin clients will fit in perfectly.

For typical corporate desktop users, thin is just right and VOA-friendly. The cost on the server side used to be high, but with the era of 1U, eight-core racks and 10 Gigabit Ethernet heading for the mainstream in 2007, there’s no question that server-hosted apps will be cheaper than desktops. Thin requires extra planning in terms of arranging offline access to data and apps for roaming users -- the fewer of these you have, the more sense thin makes -- and potentially mapping out wireless coverage zones. Stick with me; we’ll work through that.

An alternative for organizations that either need to keep firepower on users’ desks or don’t have as many seats to serve is application virtualization, an approach that Softricity is making more attractive with SoftGrid 4.0. The idea behind SoftGrid is that servers host images of individual applications rather than images of complete virtual systems. When a user double-clicks on an application’s icon to launch it, SoftGrid streams down to the client only as much of that application’s executable data as is required to get it running. Then, whenever the app reaches out for an external resource such as a dynamically loadable library, SoftGrid streams that piece, and only that piece, down to the client in real time. The kicker here is that applications are never truly installed on users’ systems. SoftGrid virtualizes just enough of Windows to create a container that tricks each application into thinking it’s been fully installed. This is far more efficient and much easier to monitor and control than launching applications from network drives, and it carries none of thin clients’ burden on server resources. And like thin clients, application virtualization is VOA-friendly because users don’t need to know or care where their applications live. And there’s a bonus: SoftGrid accommodates roaming and offline usage.

I tend to lay VOA out as a great banquet, but you might be wise to take it in small bites. The client is a good place to start.