Creating an OS virtualization architecture

OS virtualization's advantages, when hosted atop hypervisor-based architectures, fit into many different virtualization strategies. Expert Greg Shields discusses how hosted desktops, test/dev environments and servers can benefit.

Increased processing efficiency and decreased on-disk resource utilization results when virtual environments (VEs) reside within an OS virtualization environment, being stored as deltas between the host and the individual configurations of the VE. For IT organizations considering a move to virtualization, those specific differences can make OS virtualization a good fit for certain types of workloads.

In my last article, I talked about the basic structure of OS virtualization as well as some of the key ways in which it is different from the typical hypervisor-based virtualization architecture. In addition to typical infrastructure servers that make good virtualization candidates on all platforms, OS virtualization's efficiencies provide advantages to otherwise unavailable or expensive workloads when hosted atop hypervisor-based architectures. These workloads include hosted desktops, development and test environments, and high resource use servers that would not otherwise be virtualized.

Less horizontal VDI
Hosted desktop environments, sometimes also called virtual desktop infrastructure (VDI), can be a great solution for bringing otherwise highly-distributed desktops under unified management. With VDI, individual PCs are replaced with devices that connect the user into a server-hosted desktop instance. Desktops are virtualized back into the datacenter and accessed via thin network protocols like Microsoft's RDP, Citrix's ICA or Virtual Network Computing's VNC. Administrators gain from centralizing desktop processing into a small number of virtual hosts and touch points. Users gain the ability to access their personalized desktop from anywhere with a network connection, as well as other benefits like rapid deployment, snapshotting and easier troubleshooting.

But using traditional hypervisor-based solutions for hosting VDI environments can be costly in terms of resources. With each individual virtual machine consuming tens of gigabytes of disk space, horizontally scaling hosted desktops incurs a significant cost for storage. Individual hosted desktops often contain a fairly similar load of applications and configurations. These similarities are particularly effective when run atop an OS virtualization environment. Applications and configurations that are common to all hosted desktops can be installed to the host OS. Each residing virtual machine then gains the use of those resources without requiring massive duplication of their files and other elements.

Since OS virtualization is heavily focused on server management in addition to the typical functions of virtualization, an implementation of OS virtualization typically also enables a set of integrated systems management tools. These tools can be used for the installation of software, applications and configurations to hosts and VEs in addition to the usual virtualization functions like power on, power off and snapshotting.

Less tests
Test and development environments are great candidates for any kind of virtualization. These environments see a very rapid rate of turnover and change, due to the need for multiple tests and testing threads to be run simultaneously. With daily builds requiring rapid – and sometimes daily – rebuilds to the environment, the automation capabilities that come from virtualization are critical in meeting tight development deadlines.

As with hosted desktops, when organizations require test and development environments, the computers that make up each environment tend to be very similar in configuration to each other. Large in number but small in individual differences, one machine instance is often replicated many times over to support multiple tests and multiple phases of development.

Adding delta-based storage of VEs to the environment means a smaller disk footprint is required to store each test's needed machine instances. But this smaller size also improves the speed in which test machines can be turned around for use by a follow-on test. The sheer time element involved with copying large virtual machines from one part of the environment to another can impact the amount of time available to complete an environment rebuild. Individual VEs and their templates can be significantly faster to deploy, and updating the test environment to a new build gains the same benefits as with other software.

Less overuse, more resources
Server workloads that often end up remaining housed on physical resources can be reduced as well. With traditional hypervisor-based virtualization products, the resource overhead required to support driver emulation can negatively impact the ability for high resource use workloads from functioning well within a virtualization environment.

One example is with memory use. Although memory between collocated virtual machines can be shared to some extent with hypervisor-based virtualization, each virtual machine running requires a greater amount of available RAM to support its functionality than with collocated VEs atop a OS virtualization architecture. The result of this is that workloads which tend to consume high amounts of physical resources can operate with higher performance when run as VEs rather than as virtual machines.

Although OS virtualization isn't the solution for every workload, it does provide an alternate solution for IT organizations in circumstances like those above. Combinations of OS virtualization with other architectures are also worth considering, with hypervisor-based solutions driving some hosting needs while relegating others to the OS virtualization environment. As with any virtualization strategy, consider the management components available to administer the environment as critical to getting the results you need.

ABOUT THE AUTHOR:Greg Shields, MVP, is a co-founder and IT guru with Concentrated Technology (www.concentratedtechnology.com) with nearly 15 years of IT architecture and enterprise administration experience. He is an IT trainer and speaker on such IT topics as Microsoft administration, systems management and monitoring, and virtualization. His recent book Windows Server 2008: What's New/What's Changed is available from SAPIEN Press.

Dig Deeper on Introduction to virtualization and how-tos