VMware with a Modular Desktop View

a technology blog with a focus on virtualization and cloud computing

VMware with a Modular Desktop View

June 27th, 2010 |

Introduction

I come from a family of artists and I thought some of that intrinsic genetic value kind of flowed in the blood, but by my white boarding skills (illustrated above) apparently that isn’t the case. I promise I will work on my happy little trees as time permits (Bob Ross reference) and try to move away from my chicken scratch art work. I only hope that I can make both my family and Bob proud.

I have been pretty busy over the past few weeks with my new career change over at VMware. The travel has settled down a bit more for me, and I am starting to get some great quality time in front of my customers. Talking with them and listening to a lot of their challenges, has been one of the more exciting parts of my new role thus far. I have an engrained appetite to solve problems, so at times I have to consciously force myself to listen before trying to jump to technical solutions that might help them out in some way, shape, or form.

As many of you know virtual desktop solutions are the talk of the IT industry right now. People are out evaluating what all the hype is about, and what’s so special about the various offerings in the industry. I wanted to share some of the VMware concepts with this blog post because I think there is some misunderstanding of how VMware View works in combination with VMware ThinApp and what we are trying to accomplish. I also want to discuss why some projects are stalling, and how you can protect yourself, and in turn overcome some of the common roadblocks and obstacles along the way.

Desktop Virtualization

So what is VDI (Virtual Desktop Infrastructure) with VMware? Virtualizing desktops is the same thing we have been doing to servers for years. In its most basic form, take the end user desktop and abstract the operating system from the hardware. Provide a more robust platform (aka a server) to run the desktop on which will then offer a more robust end user experience. Integrate the powerful platform of VMware vSphere to leverage such things as VMotion/DRS/HA, while consolidating many desktops to one physical system. Wrap a robust web based management interface around this solution with a rich desktop protocol (PCoIP) and this is what we call VMware View.

Application Virtualization

Why stop at just the desktop operating system? VMware can take you one step further in your quest for an elastic IT infrastructure with VMware ThinApp. Virtualizing your applications can now eliminate some of the most common helpdesk tickets placed by your end users. Break the bond between the application conflicts and their pollution into the operating systems space. Run them off a central file server which eliminates the need to distribute large cumbersome application updates to all of your end users. Encapsulate older legacy applications that were designed for older versions of Microsoft Windows and allow them to now run on Microsoft Windows 7. This is an excellent way to help speed up your migration path off Windows XP!

Combining these two technologies can create a more of an attractive solution for you and your IT organization. Why should you have a mapping of users to specific desktop configurations? If you’re like me, your thinking “because they need access to their applications to get their jobs done”. Bingo! Abstract it by virtualizing it, then create common pools of resources that allows IT to run more as a service, not a customized solution that creates more work for your organization. Think about it.

Sounds great, where do I start?

A lot of feedback I have been getting around VDI (Virtual Desktop Infrastructure) is that people are ready for this next big change in their IT infrastructure. Some have dabbled in it, some are using it to a degree, and some (not many) have rolled massive deployments out to their end users or customers. The end result is that people know they need it from a fundamental perspective to make life easier for both the IT organization as well as the customers they support. As with any major change in IT infrastructure there are several reasons why these projects are successful and why they aren’t.

One of the biggest reasons we see desktop virtualization projects stalling out is because some of the same assumptions of server virtualization are made and applied to the desktop infrastructure. If you take away nothing else from this post, take this with you.

Virtualizing desktop workloads is a completely different animal!

Take a step back and think about server workloads compared to desktop workloads. Even the most unpredictable server workloads have a degree of predictability. You normally find a limited number of applications, or in many cases one main application per server. That application has known metrics tied to it on how it is expected to act and behave, how it will utilize CPU, memory and storage. System backup’s typically occur at the same time every night, or at least within the same window of time. Virus scanning is a scheduled activity normally set by the systems administrators on each server.

Turn the table and consider the desktop user environment in comparison. Despite how hard your desktop support team attempts to maintain one single image, there are many different system configurations for various groups within the organization. There are typical end users, power users, mobile users (sales force), VIP end users (C Level) and the most difficult of all, the IT end users. Many organizations let their end users install any applications on their desktops or laptops which can result in conflicting software installations. Virus scan’s typically are launched on a scheduled interval, which effects all of the desktops within an organization. Throw human factors into the equation now, users playing games, users watching YouTube video’s, users going on vacation, working from home, working after hours to meet project deadlines, coming into the office early or maybe staying late? Talk about unpredictable, it’s no wonder why some VDI implementations end up looking like this guy!

VDI Planning and Assessments

Don’t let me scare you off. I didn’t say this next step for you is impossible, it’s just a little different and requires some leg work. As you begin to look at desktop virtualization, do yourself a favor and do some up front planning to ensure you are successful. I am going to try to make you look like an expert when it comes to this new fangled VDI stuff. Start by forming a VDI project team which consists of the right mix of people within your organization. I would propose the following:

Management sponsorship (fund the project)
Project Manager (drive the project)
Systems and Network Engineering (Infrastructure requirements)
Server Virtualization team members (Manage the VMware infrastructure)
Desktop Support team members (Mold and support the new VDI environment)
Key stakeholder from each end user department (Validation and buy in from each respective business unit)
Application support team (Application guidance if embracing VMware ThinApp)

Once your team is formed and you have a fundamental approach of your own VDI implementation, I would highly recommend doing a desktop assessment. Many organizations think they know the end user environment and what their needs are. I would challenge that thought, organizations are constantly changing. Business demands change internal organizational cultures constantly and there is also the consideration of people leaving the company or being promoted into different positions.

There are two popular tools that will help you asses your physical desktop environment and help you migrate to a virtual desktop configuration. VMware recommends and supports both of these products.

Check out Lakeside Software’s Systrack application. I have actually done some training with Lakeside and really like what I see. I have 10+ current active assessments going on right now, and the software does an amazing job with helping both understand your current environment and then mapping that to what you need from an infrastructure perspective. They also provide a great power management application that can help fund the entire cost of the software.

The other popular tool that you should consider evaluating is Liquidware Labs Stratusphere. Stratusphere utilizes a virtual appliance for it’s centralized administration in your environment. It also takes the same approach in pulling data from your existing configuration and producing some great reports to help guide your architectural decisions.

Conclusion

Hopefully this helped to explain what VDI is if your new to the concept and how VMware’s approach to this is different than most others. Let me know where you are in your VDI/View deployment and what you have found to work or not work!

Pingback: Tweets that mention VMware with a Modular Desktop View | virtual insanity -- Topsy.com
http://seanclark.us vseanclark

Awesome post, Scott! Please consider turning this post into a static page that is easily linked to for folks just beginning their journey into VDI. I agree with a ton of this. Creating the right VDI center of excellence or team is spot on. Assessing your environment with tools like Liquidware Labs Stratusphere and defining good requirements for the VDI project is also huge. Good work!

But, I have to add my $.02 so I hope you don't mind. You project team is missing a key area. If the company has a a Terminal Server team in-house, you'll want to make sure you involve them because they've been bleeding on similar issues for years and their experiences map pretty well to the VDI space.

And if I wanted to get nit-picky, if the company is big enough to have a dedicated storage team, you need to involve those guys so they can start to begin thinking out of the box (if they aren't already) on the unique storage demands that need to be addressed for rounds 2 and 3 of VDI deployements.

And remember, with VDI we don't make mistakes, we just make happy little accidents…..in the POC.
scottdsauer

Thanks Sean! I agree a core “VDI A-Team” is critical in having success with these types of projects. The Terminal server or Citrix team is also another important team to have in the mix, thanks for picking up on that one. I actually was thinking of the storage team when I was writing the post, but mentioned them as “Systems Engineering” since that was my old position title! I completely agree with your comments around storage deployments and VDI. The requirements are unique and should be closely monitored not only at the start of the project, but throughout the life of the VDI deployment.

Thanks for taking the time to post your feedback, as always it's much appreciated!

Scott
jjj

But to be fair…

– Both VDI solutions (the two real competitors) charge $180-$200 per user for the “privilege” of running VDI. This does not include the cost of the storage, server hardware, desktop OS license, client to access the virtual desktop, solution for accessing the desktops remotely (SSL vpn, etc.). That's a hard number for customers to handle.

- All ROI calculations, if and when they are presented (which never seems to happen. Prove me wrong and link one, please), require cutting staff to make up for cost (vdi = initial outlay of $$ to save management $$ later, right?). This is usually the same staff that participates in the VDI rollout.

- Both real competitors have serious flaws. XenDesktop requires XenApp in order to deliver a complete solution (which begs the question, why bother with XenDesktop at all? Why not stick with XA?). View has no realistic WAN solution at present (and no Windows 7 support, and no way to re-provision old desktop hardware).

In order for VDI to go from Nice-Idea-That-Doesn't-Make-Any-Financial-Sense-Except-In-Very-Rare-Cases to a legitimate solution like you suggest, I think VMware needs to do the following (note that I'll Citrix out of the discussion, but number 1 applies to them too)

1. Drop the cost to something reasonable. $100 per user is more palatable.
2. Drop the notion that the ESX licenses purchased with View are only allowed to be used for desktops (it's time to let people mix desktop & server loads)
3. Fix the WAN problem (RDP is not a real option)

IainR

Hi there, we're deploying this into several customers at the moment, and in response to your points, the cost is per concurrent access (as opposed to per named user in the XD world), and in the majority of cases, our customers don't want to mix infrastructure to manage desktops and server loads. In any case, they can just purchase vSphere licenses and the Add-on View licenses, then they are ok to mix the environment.
In every deployment so far, we have a very small number of RDP based users, and the rest (even on LAN) use PCoIP. We have quite a few heavy graphics users with Teradici thin clients to support end point acceleration.
Have a play, there's a lot of “old” arguments here that have been solved. Enjoy!

scottdsauer

Thanks guys, good points but I do agree with IainR. With over 1.5 million seats of view sold, we are beyond a nice idea and are now into prime time production.

jjj

That number is nice, but let's be fair, it's small relative to the base installs of traditional desktops that VDI vendors are targeting. Companies like Parallels can boast those kinds of numbers (can't they?) and it would be hard to think of their solution as being prime time. Plus, a large portion of the 1.5 mil sold would be health care, where large seat counts enhance the licenses sold number, would it not (and where View IS a great fit)? And “sold” is not “implemented”.

But your point is taken, “rare” may not be a fair way to categorize this. Is it fair to say niche for now?

jjj

I did not know about the option to go with add-on View licenses, so I stand corrected. I do think that having this option to mix loads – though I totally agree that in anything more than a small scale this “mixing” is a bad idea – is the way to bring in SMBs and even enterprise companies who want to try on a small scale before committing (an easier thing to do few good ROIs exist). Sort of the idea of “let's roll out View….to the IT dept and kiosks / low-use users first” before we commit the bank).

I also understand that RDP is not the play here with View, it's PCoIP. But WAN performance – or maybe better put, Wan scaling, is still a real concern. If you're rolling out PCoIP on the WAN (not LAN, where yes, it does do nicely), are you going with Security Server today, SSL VPN / load balancer tomorrow? Or starting these roll outs with hardware appliances in-place? How do you combat the WAN (not LAN) performance of ICA? (note: not a pro-XD ranter, I am well aware of the things I like better about View, like how much easier it is to setup, how I don't need a physical server, how I have ThinApps versus “I'm still stuck running XA for app delivery)

And some of the “old” arguments are old, but not yet over, like 7 support (which I know is coming very soon. Not here yet though) and ROIs for the masses (maybe that's a better way to put it than no ROI at all. My initial post, on re-reading, comes off as a bit more rude than I meant it). Others I didn't mention include a lack of controlled bandwidth throttling (except printing), a relatively low cap (less than 5K) on concurrent sessions per server, no redirect of client drives, limited USB device support (aka cameras), no OpenGL or DirectX….and more.

I get that it's a work in progress too, so it's a bit unfair to expect all these things to be working already – some of them haven't been solved by competitors (and competing products) that have been in the market for far longer. But with 4.5 expected to change a good deal of the design, and these concerns I raise, I still think this is a rare fit for now with a promise for greater appeal to the masses – maybe, if problems are addressed and costs are dropped – in the future.

Cheers

scottdsauer

Kelly,
Yes it's hard going back and retrofitting a project that has already had some traction and then they hit the brick wall for various reasons. My hopes with this post was to give users visibility of some of the upfront challenges and try to start off with an assesment to help understand what is needed.

Thanks for the comment!
Scott