Cost and graphics concerns delay a VDI project

Virtual desktops are a serious paradigm shift and Scott Lowe is taking it in a slow and measured way. In this article, Scott provides you with an update on ongoing VDI efforts at Westminster College.

I've written previously about my efforts at Westminster College to displace the current desktop replacement cycle in favor of a solution that costs less, is more efficient and provides mobility to both the workforce and the student body. In an ideal world, while I doubt we'll eliminate our physical computer lab spaces in the near future, I'd love to augment these limited access (16 hours per day, on campus only) spaces with an always-available service accessible from wherever our students, faculty and staff happen to find themselves.

We are moving very slowly on moving forward with anything beyond a lab implementation while we continue to work on making sure that whatever we implement ends up being well-accepted by our audience.  Specifically, any virtual desktop project needs to mimic as closely as possible the service and performance provided by a regular client. We're not expecting to implement VDI for services that require major graphics processing, but do need the VDI solution to reasonably replicate performance for normal office duties and web browsing habits, to include relatively significant use of Flash video and audio.  The Flash audio side of the equation is pretty good, but video leaves a whole lot to be desired. As time has been available in the past few months - and there hasn't been a lot - I've tested different solutions to the Flash issue which I think is really the only remaining hurdle to a more general pilot test.

I've been doing Flash testing with Wyse's V10L line of pretty inexpensive terminals - to no avail.  To be fair, I've been fortunate to work with some Wyse engineers who have been extremely helpful and have shared beta versions of their TCX extensions.  All that said, I came to a realization - without some semblance of reasonable graphics processing capability at the thin client, there is simply no hope of replicating the desktop experience.  This probably should have been evident from the beginning, but hit me one day while I was doing testing.

While I was at a conference a couple of weeks ago, I was able to spend some quality time with VMware View people and Wyse people who recommended that, rather than the V10L terminal I'd been using, I might be better off with a C-class terminal, which I have yet to test, but do have sitting in my office waiting for an opportunity. The C-class clients include a separate graphics accelerator for much enhanced video performance.  I'm relatively confident that the C-class terminals will perform better than the V10L, but there is one serious catch: cost. The units can cost upwards of $400 each. Frankly, I can buy a full PC for that price. Of course, the VDI solution does help to meet other desires, such as enhanced mobility.

Ideally, the VDI solution will have an initial cost that doesn't significantly exceed that of a traditional solution.  I realize that the cost breakdown for VDI will break down much differently than it does for traditional desktops, with a strong likelihood that initial costs will be somewhat higher but ongoing costs coming in lower.  The lower costs will be due to lower energy usage and a longer replacement cycle for the thin clients themselves.

Even though ongoing costs will be lower, that initial investment is a tough pill to swallow, so I'm continuing to look at alternatives. While roaming around Best Buy a while back, I set my sights on a $199 full PC from Acer, called the Acer Aspire Revo.  With a 1.6 GHz Atom processor, 1GB RAM, a 160GB hard drive, and a wide variety of ports - including USB, eSATA, HDMI and memory card slots - the Revo just might be a viable alternative to the usual thin client. The Revo ships with Windows XP Home and is upgradable to up to 4GB of RAM and can be outfitted with an 802.11n wireless adapter.  Further, the Revo includes an Nvidia ION LE graphics adapter.  Although it's a shared memory graphics solution, as a terminal, that shared memory aspect shouldn't be a major hindrance.  Adding an additional 1GB of RAM is very easy and very cheap and the Revo is supposed to be upgradable to Windows 7, a claim that I'm working on testing.

As I get further into testing the Wyse C-class and the Revo, I'll report back with results.

Now, on to Windows 7. Windows 7 includes a significantly enhanced version of RDP that is designed to improve a number of features that make it easier to deploy virtual desktops, including support for real bidirectional audio, multiple monitor support and enhancements related to the display of multimedia. Reviews of RDP 7 on its own have been favorable with regard to overall performance, so I look forward to testing it for myself.

Further, with the release of VMware View 4, VMware has integrated Teradici's PC-over-IP (PCoIP) service into the product.  PCoIP is a remote graphics protocol that VMware has turned into a software service capable of server-based complex graphics rendering that is sent down to a client.  PCoIP tunes itself to provide the best experience possible given inherent restrictions, such as bandwidth availability. On the flip side, as a server-based service, I can't see how PCoIP can actually work without reducing the clients-per-server ratio.  After all that extra processing power has to come from somewhere.  Fewer clients per server means that more servers will be needed to support the virtual desktop infrastructure.  That could break the cost model, but servers aren't really all that expensive anymore, so it's still worth investigating.  I should note that Wyse has also announced that they will support PCoIP in many of their thin clients, so it might be time to retest that V10L to see how (and if) it operates with PCoIP.  Wyse has released a new client - the P20 - solely for the purpose of using PCoIP and the P20 contains hardware specific to PCoIP. However, the price of the P20 far exceeds that of even the C-class that I mentioned earlier, so it may not be a good fit from a cost perspective.

I'm continuing to research, test and learn about all of the ins and outs of VDI, and want to make sure that, if we go down this road, we do so with the right information and that we know exactly what we're getting into from a user experience perspective. I remain confident that VDI can and will save money over time and that, given enough time, we'll be able to solve the user experience issues.

If you have any tips about VMware View 4, the Acer Revo, RDP 7 or PCoIP, I look very forward to your comments.