For CIOs, the distributed computing model with a fat client at a user's fingertips has long been a cost, deployment...
and administrative nightmare. But is the devil you know (fat clients) better than the one you don't (virtual desktops)?
With fat clients, users get a familiar operating system, real-time access to multimedia applications and device portability. On the other hand, end users often end up trying to fix their own PC problems and add programs that may cause the device or even the network to go haywire. Laptops loaded with a lot of data pose serious security risks.
With desktop virtualization, security risks are minimized since the data resides in the data center and not the client. It's easier for IT to send out patches and updates and perform backups from a central location, and end-user downtime is decreased because a worker can just jump on another device if his system crashes.
Yet with virtual desktops, users have to be connected to the network at all times, there is limited driver support for virtual desktop infrastructure (VDI) machines (more on VDI later) and users will complain about the performance of graphics-laden applications since such applications no longer reside locally but have to travel over the network.
As companies with successful server virtualization efforts under their belt begin to look at desktop virtualization as a possible alternative to fat clients, the question of what technical approach to use, plus how to prove the business case and mitigate any end-user grumbling, are top of mind for CIOs. Indeed, in a survey conducted this month with 93 respondents, SearchCIO.com found that 24% of respondents expect at least limited use of desktop virtualization in 2009, while another third plan to evaluate the technology this year. Some 40% had no plans to explore or use desktop virtualization this year.
Models for desktop virtualization
VDI is the most commonly known model for desktop virtualization. With VDI, IT virtualizes desktop images and stores them on servers that then stream them over a network to end-user devices. This scenario cuts costs by centralizing desktop images and associated desktop management tasks, but it also opens the door to new storage and network infrastructure costs.
There are other models on the horizon, including bare metal client hypervisors being developed by VMware Inc. and Citrix Systems Inc. that would eliminate some of the drawbacks of VDI. Bare metal would give users offline access and IT a better balance between desktop management and flexibility for users in their choice of devices and applications.
With the VDI model, all of the computing power that was once on the desktop is now located in the data center. "The reality is there's not a whole lot of companies that have spare capacity in their data center, so if you move to VDI it often means increasing the size of the data center or adding capacity to it, and it's not trivial by any means," said David Payne, chief technology officer at virtualization integrator Xcedex Inc. in Wayzata, Minn.
When Xcedex works on a server virtualization consolidation project, vs. a desktop virtualization consolidation project, the team often budgets more computing power for PCs than for servers. For CIOs who then want to virtualize desktops, this translates into not just the cost of developing a new desktop deployment strategy, but also buying new servers and server software.
But storage is the most overlooked cost of a virtual desktop strategy and could be the biggest cost delta in moving to a desktop virtualization environment.
"Oftentimes a virtual desktop strategy means buying SAN storage, which has a host of other costs associated with it," Payne said. "You need to put the SAN infrastructure in place and the network needed for that and relocate all the data, which is now on distributed low-cost drives today, and take that capacity into the data center."
The network implications can also be substantial when you no longer have the computing power on the fat client but instead have 1,000 users conducting 100 MB remote sessions to the data center per second. That bandwidth usage must also be continuous, compared with fat clients, which communicate with servers in spurts such as when a user downloads a file from a server or writes data to the SQL database. CIOs need to figure out if the company's WAN links are capable of handling that type of traffic.
A good way for CIOs to approach added network and storage costs is by combining the buying power of various projects under way. Large enterprises often realize too late that there is overlap among server consolidation, desktop virtualization and data center virtualization tactics.
"On the VDI side, say if you're using VMware or Citrix [technologies] as a VDI engine … well, that hypervisor infrastructure can be leveraged for servers as well as desktops," Payne said. "If you need 20 hypervisors for server consolidation and 20 more for desktop virtualization, that is a perfect time for the CIO to make one big purchase and save some money."
The inability to be untethered from the network to use a virtual desktop is a big hurdle CIOs must overcome compared with fat clients, in which a user's personal setting and applications are housed locally.
This leads to the biggest drawback of desktop virtualization versus fat clients: end-user experience. Not only can users not work offline -- although this is an area vendors are working on -- but the model is also simply not the right fit for many types of users.
The reality is there are not a whole lot of companies that have spare capacity in their data center, so if you move to VDI it often means increasing the size of the data center or adding capacity to it.
David Payne, chief technology officer, Xcedex Inc.
"End users are saying [VDI] is not good enough, it's too slow [compared with a fat client] and this gets back to the department head, then to the VP and then the CIO. End-user pushback is a big problem for CIOs," said Andi Mann, an analyst at Enterprise Management Associates Inc., a Boulder, Colo., consulting firm.
To reduce end-user pushback, CIOs can pilot desktop virtualization with departments or groups of workers who perform the same daily tasks. Call center workers will not balk as much as mobile power users if they lose some applications and personal settings and their workstations are locked down.
"I see enterprises piloting desktop virtualization projects in remote offices where that office is having some issues around desktop maintenance and management -- such as having to wait days for someone to visit and fix their PC," Mann said. "That's something that desktop virtualization can resolve quickly."
Proving the ROI through a departmental or branch-office approach may also lead other departments to agree to pay for their own virtual desktop deployment. The companies having the easiest time as far as buy-in for such projects are enterprises that have centralized IT budgets and teams working together, said Mann, who wrote a white paper on the impact of virtualization on business users.
CIOs need to reach a balance: give users a virtual desktop experience that is as good as or equal to using a fat client, while meeting IT and corporate demands for better management, efficiencies and security.
Let us know what you think about the story; email Christina Torode, Senior News Writer.