Institutional Desktop Virtualization Efforts - Any Thoughts?

Institutional Desktop Virtualization Efforts - Any Thoughts?

There are a lot of technical posts here so I thought I might bring up a question about practical applications of virtualization. I have worked at many universities as an IT worker and contractor and have seen a large adoption of virtual desktops. One of these, the University of British Columbia (UBC), has been engaged in an ambitious program of desktop virtualization.Virtualization of the computing infrastructure at the University of British Columbia began about 2005 with an effort to better manage the storage grid ( and continues with a virtual desktop programbased onVMWares View 4product ( expected benefits of the virtual desktop project are:

  • Improved Data Security - elimination of data islands.
  • Standardization of desktop platforms and support by adoptingVirtual Desktop Infrastructure (VDI).
  • Using thin-clientsto increase hardware life-cycle and reduce power costs and CO2 emissions.
  • Simplify and facilitate campus-wide license arrangements andIT Service Management (ITSM) program.
  • Remote desktop access.

An added bonus is other savings that IT restructuring brings like printing savings by creating aninstitutionalplatform for hard copydigitization and document management ( here is the question: Is there anyone in this forum working on this or similar virtualization efforts and if so, what are your results, problems, observations, suggestions?

7 posts / 0 new
Last post
For more complete information about compiler optimizations, see our Optimization Notice.

So here is the question: Is there anyone in this forum working on this or similar virtualization efforts and if so, what are your results, problems, observations, suggestions?

As a matter of fact my team does work on server virtualization analysis in general, and VDI is one of the workloads we do look at. I think one of the problems is the tremendous "variance" in usage models within this area. From doing graphics rendering on the server, to offloading it to the client, and from power users to "internet/email" users this is definitely not a one size fits all workload.

With that said, the point we are at now is trying to understand the workloads' [many] demands on the actual virtualized server, understanding the bottlenecks, processor needs, etc...

I will try to write a blog post about this in the near future.

For now though, I am curious about your personal experience with this workload if any?

Tom referenced desktop virtualization and you server virtualization. Are you both asking the same questions, or is Tom's question going in a slightly different direction? Does it matter?

Using thin-clients to increase hardware life-cycle and reduce power costs and CO2 emissions.

Tom, have you looked at:

VDI stands forVirtual Desktop Infrastructure, the idea of using a server to offload some standard desktop processing tasks, and then using something like Remote Desktop to access it.

From the server's perspective, its "just" a workload, but is interesting because it can involve a very large number of VMs with relatively low levels of CPU utilization for each.

What I was trying to say was that my team focuses on studying the server side of VDI and understanding how it affects CPU, I/O, Memory, Caches, etc...

personal experience in setting up and using a virtual system is mainly with application
software running on a server and users accessed the system using a web browser
and Citrix through a VPN. The system supported several hundred simultaneous

biggest bottleneck was the network itself. At locations that had a very high
speed/capacity network infrastructure the screen update was almost as fast as
working on the server itself. Where the network was not as robust the users
constantly complained that anything they did took several seconds to respond. Aside
from the network lag time there were few issues and using a virtual interface
solved a great deal of administrative problems and reduced costs. The software
only had to be installed and updated at one place. The installation of the
software (an EPR/CRM) was complicated and required extensive customization and
an Oracle client needed to be created and setup on a user's machine if the
software was being run on it, which again added to the complexity. Also, many
users owned Apple Macintosh laptops and the software was C++ WIN32 based and
buying Windows and adding Apple Boot Camp or running a virtual machine was not
a cost effective solution and would add to technical support loads.

Other added benefits were:

Enhanced security - users did not require a
database password on the personal computer

Remote user access - schedules could be checked
from home

Remote Technical support - a technician could
run the system and see the problem first hand

I looked up the bandwidth consumed by XenDesktop and it varies
greatly. An Office app will be about 43 kbps while hi-def video runs 1812 kbps.

takes up a large part of thebandwidth consumed (553-593 kbps) due to the
data that has to be sent to the desktop machine. I would consider disabling any
printing functions if possible. The network latency is usually not created by
many users using the virtual desktop but by all of the other network traffic
combined with it. Often the only solution to slow response times is to upgrade
the entire network system when implementing a large virtual desktop solution.

Very nice (and FREE!) SDK.

Leave a Comment

Please sign in to add a comment. Not a member? Join today