Real world virtualization: realizing the business benefits of application and server virtualization; Dealing with virtualization challenges.

Do you believe virtualization is the answer to squeezing more value out of IT? Depending on whom you talk to, it is either over-hyped or the silver bullet. Like most advanced technologies, the answer lies somewhere in between. However, you cannot deny the strong growth of virtualization technologies. Already close to 100 percent of the Fortune 1000 use some form of application virtualization. Indeed, aAccording to IDC, seventy-five percent of all companies with 500+ employees are in the midst of deploying virtual servers. While there are multiple virtualization technologies available, I will focus on application virtualization technologies such as Citrix Presentation Server and server virtualization technologies such as EMC VMware. Whether you are a newbie to virtualization or are knee deep in it, this article should help you understand--and more effectively deal with--the challenges you are experiencing now or could encounter in the future.

Application Virtualization

Application virtualization is the new fancy trendy name for server-based computing. However, instead of installing applications on desktops, the applications are installed in a server farm for secure, remote access. Server virtualization allows you to take multiple physical servers and create the same number of virtual servers, or "machines," on one host physical server.

For a first-hand perspective on how to prepare for and overcome challenges with application virtualization, I turned to Ray Leitz, CTO of AcXess, a business continuity provider for SMEs. Leitz was formerly CTO at a Fortune 150 company where his group managed a Citrix implementation with 18,000+ named users (4,500 concurrent users), 300+ published (i.e., virtualized) applications, and more than 170 servers.

According to Leitz, the biggest challenge is in dealing with two inevitable paradigm shifts. "Your administrators and end-users each will have new realities. Administrators are accustomed to supporting desktops where a poorly installed application only impacts one user. Now they are installing an application on a remote server that is published to potentially hundreds or thousands of users. This means they must be much more disciplined in testing their deployments.

"In the case of end-users, they are used to having control of their desktops. Now their desktop may be replaced by a thin client and they experience a loss of control, not understanding how this technology will benefit them."

Because of these shifts, some expectations need to be set. "For the end-user, there are three inherent benefits you can sell them on," says Leitz. "First, in this new virtualized world the end-user can now work securely from any device/location, including home, instead of working late at the office. Second, desktop hard drive failures are no longer catastrophic because the end-user can acquire another desktop or go somewhere else with web access to begin working again. Third, application virtualization also takes the burden of backups off of the end-user. Even if the end-user had a backup, it could take hours to restore, costing them productivity."

Selling the end-users before rollout makes sense: once they understand these benefits--beyond saving the company money--they will be more tolerant when problems arise.

You don't usually have to sell the administrators--do they really have a choice? However, there are some best practices that can help them be more efficient and effective once deployed. Leitz found that there were three key factors in maximizing their application virtualization investment.

1. Prepare for the worst. With application virtualization and proper planning, you can take advantage of inherent business continuity protection; such as if a server goes down, you can quickly swap in another server to handle the load. "It's important to add excess capacity in case server(s) go down," says Leitz. "This avoids impacting your SLAs." However, also be sure to have a continuity plan if your data center is impacted by a disaster. Leitz continues, "Although your users can access from any remote location, you still need to ensure that the data center itself is not a single point of failure."

2. Focus on the four A's. A--authentication, authorization, audit, and administration are even more important because a centralized environment can open the door to internal abuse of privileges. Use the four A's to segment employees by role, location, and department. "If you have good security and administration policies in place, you can make it very difficult for people to abuse the environment," says Leitz.

3. Don't overlook reporting. Beyond verifying compliance and security policies, there are many other reasons to have a robust reporting solution. In this centralized environment, you should be able to monitor usage trends such as total cost per user, over or under utilization, and licensing compliance. "The money for building a central server farm can be a big target for the CFO," says Leitz. "Counting the costs of licensing for infrastructure, hardware, and support staff, application virtualization can result in an average savings of 40 percent or more in per-user costs per month."

Leitz is quick to stress the critical need for a sufficient, targeted pilot rollout. He suggests including middle managers as a pilot group because they can complain the loudest if their departments are impacted by an unsuccessful rollout. "By getting their buy in and seeing that their key applications are performing acceptably, you will have a better picture of what the real ultimate deployment will look like and can plan better for it," says Leitz.

Finally, Leitz suggests carefully evaluating the impact of application virtualization on your SLAs. "Having a new employee up and running should go from a couple days to hours or minutes," says Leitz. "You can also move employees faster because their access point is irrelevant. And although version upgrades will take less time to deploy, be sure to lengthen time for brand new application deployments to allow for more upfront testing."

Server Virtualization

For insights into challenges in the server virtualization world, I spoke with Dave McCrory, an independent virtualization consultant and co-author of "Advanced Server Virtualization."

According to McCrory, management of virtual server environments can be extremely daunting. "There is no comprehensive single management tool for planning and troubleshooting because most tools have no or very little awareness of virtual machines," says McCrory. "This is because they weren't originally built with virtualization in mind, but instead are being adapted for virtualization management. To make things more challenging, virtualization technologies are still evolving into next generation hypervisor environments, which, instead of residing on top of Linux or Microsoft Windows, have their own operating system on bare metal hardware."

There are some best practices that can help you get a handle on this fast growing and changing technology. McCrory provided these four best practices that will allow you to avoid common pitfalls.

1. Standard conventions. Standardize on a method to track machines and what resides on them to better organize your virtual assets. "The more you can standardize, the easier it is to look at one physical host and know exactly what is installed there," says McCrory. "This will help you better manage your virtual assets and avoid shutting down a server and unintentionally impacting more users or applications than you expected."

2. Test thoroughly. Seems like common sense, but it makes it no less important. McCrory suggests starting a pilot with IT because they will be the most tolerant subjects. "A pilot gives the IT group a chance to encounter issues the eventual end-users will also come across," says McCrory. "And it gives your administrators the chance to practice and get comfortable with this technology, while serving a friendly audience." McCrory goes on to say, "When you do roll out production applications, I strongly recommend not starting with more complex applications that stretch limits of virtualization. A bad choice would be an Oracle database with 250 gigs of storage with millions of daily transactions. Instead, stick to something with a smaller number of users and a lighter weight application."

3. Track performance. Find an information gathering solution to track what machines are doing what, grouped by virtual machines and their physical hosts. If you have the budget and skills, you can build something custom, or just start with the vendors' monitoring solutions, such as VMware's Virtual Center. "It is valuable to understand your virtual and physical machine utilization trends," says McCrory. "If utilization on physical machine is only at 30 percent, you can add more virtual machines without incurring the extra expense of adding new equipment."

4. Leverage savings opportunities. There are cost savings opportunities to be had here as well. McCrory suggests leveraging your virtual machine savings potential in support of your SLAs. "You can get more savings in high availability because one physical server with four virtual machines can serve as a fail-over for five physical machines, saving you hardware costs." Also, consider how a virtual machine's transient nature could result in a software license being consumed, at which point it's decommissioned. "It is not easy to manage the "virtual" consumption of licenses, so it is best to thoroughly review your license agreements to see if you are bound by physical CPUs," says McCrory. "There may be hidden benefits, but if you aren't sure, it is best to have a discussion with the vendor."

McCrory has seen many of his clients try to virtualize everything early on, only to find some applications and machines are not good candidates for virtualization. Some make the mistake of continuing to struggle and try to make them work, only to compound the problem. "It is best to benchmark what machines are currently consuming so you can decide which physical servers are good candidates to host virtual machines," says McCrory. "Finally, keep in mind that a hypervisor virtual server runs with about .5 percent overhead compared to the 12 percent overhead in a non-hypervisor environment."

To sum up, application and server virtualization offer challenges as well as a myriad of benefits. If you follow standards, thoroughly test through a pilot program (making sure to identify appropriate test subjects), and continuously analyze your virtual environments, you will surely experience fewer "virtual headaches" ... and actualize the real benefits that virtualization offers.

Eric Spiegel is CEO of XTS. which provides reporting and analytical software for Citrix and other virtualization platforms. (Ellicott City, Maryland).

www.xtsinc.com