Is VDI finally cheaper than traditional PCs for CapEx?

Listen to this podcast

One of the promises of VDI has always been that it's "cheaper" than traditional PCs.

One of the promises of VDI has always been that it’s "cheaper" than traditional PCs. (Actually, that's a promise of server-based computing in general, not just VDI.) Anyone who's looked into this quickly learns there are two kinds of "cheaper:" There's "cheaper to buy," a.k.a. "capital expenditures" or just "CapEx," and there's "cheaper to operatate," or "OpEx." VDI has always been the latter. In other words, VDI was sold as being cheaper to own, maintain, and operate when compared to traditional PCs. The upfront purchase cost was definitely NOT cheaper than traditional PCs (and in many cases, VDI was actually more expensive).

Of course there are a lot of reasons why people use VDI. Sure, some folks are purely cost-driven and want the cheapest solution possible. But others choose VDI for one of the other benefits (like the enhanced security or performance for three-tiered applications), so for them cost doesn't matter. (Sure, if VDI also ends up being cheaper, great, but that's not why they bought it in the first place.)

But the general consensus of the industry over the past few years has been that the savings associated with VDI come in the form of lower OpEx, not lower CapEx.

So you can imagine my surprise when I heard a customer on a panel at an event last month say that he was thinking about VDI for its CapEx savings!?! The customer explained, "Look, I have 400 users who each get a new $800 desktop every four years. So that $320,000 in desktop hardware alone. If I can build a VDI solution for under $320k which reuses the existing desktops, then I've come out ahead just with CapEx alone. Then throw in all the OpEx savings I can get like easier patching and a reduced need to drive out to remote sites to fix desktops, and this decision really becomes simple!"

This customer makes a good point!

It's important to clarify that we need to make sure people don't switch to VDI just because it could be cheaper to acquire than traditional desktop PCs. VDI has its specific use cases and well as some very serious limitations, so having a solution that's cheaper but doesn't meet the business needs doesn't make sense. (If it did, we'd all use pens and paper instead of computers.)

In the case of the customer on the panel, he's already a Microsoft virtualization user (he has no VMware and only two XenApp servers), and about 80% of his servers are running on Hyper-V 1. So he's just planning on using Microsoft's VDI bundle which he can get for $21 per device, per year. (He's also planning on using persistent images.)

So is this an isolated case of crazy? I thought so at first and just sort of filed it away into the "interesting" category and planned to revisit it at some point in the future. But last week my interest in this topic was piqued again when VMware claimed that View 4-based VDI solutions can actually have lower acquisition costs than traditional PCs. The numbers they through out were $600-$800 to for the hardware & software needed to acquire a traditional PC versus $750 for a VMware View-based desktop. (I haven't seen these reference architectures yet so I don't know how realistic they are, but the main point is that people are starting to talk about VDI as cheaper to acquire where a year ago they weren't.)

Obviously everyone's situation is different, and how many users a certain piece of server hardware can support will vary drastically based on app loads, protocol requirements, etc. But of course traditional desktop computer needs will vary too. So what do you think? Is this true? Is there something we're forgetting?

Again, it's important to remember that we're not trying to say you can ignore all the other parts of the conversation when you evaluate VDI versus traditional PCs. It's just that the "PCs are cheaper to buy" part of the argument might not be as valid soon?

Join the conversation


Send me notifications when other members comment.

Please create a username to comment.

Most of my customers pay less than $500 for a corporate PC today.  In their cases, there's no CapEx savings for VDI for the "typical office worker" yet.


IMHO there are no cost savings in either capex or opex. There are agility, security and mobility gains but they do come at significant cost.

1. VECD - If running Windows desktop OS

2. Data Centre compute resources

3. Enterprise grade SAN storage because of the read/write % and transactional/non tolerence nature of desktop style workloads

4. Hypervisor and/or Hypervisor management costs

5. Bandwidth/network accelerator additions

6. Brokering solution

7. Backup software & tapes/VTL solution

8. User personalisations type software (think AppSense etc)

and more.....

And you still need to maintain a platform for managing your "desktops"....they still need an OS of sorts.

So where are the cost savings??? Am I missing something?


The ROI or cost savings for CapEx will only be positive or neutral in a scenario :

1 - where the environment/use case is simple enough that your technology stack can small as in the case above

- or -

2 - where you have taken a holistic approach to application and desktop delivery (ADD) - VDI/HVD in particular, and been able to find budget in other areas e.g. such as any potential savings that might come from application rationalisation and the associated licensing, hardware and packaging costs that might be driven out by the rationalisation programme - but again this is only really compelling in environments of scale.

Over time it may become CapEx cost effective vs traditional PC's - we're already hearing the claims of more VMs per core with the new processors - but this is still way behind Terminal Server based solutions - which actually is still the most cost effective solutions for most requirements.


I see this a lot as my company focuses on selling both VMware and Citrix solutions.  Most of the time, we see a break even on CapEx and huge benefit in OpEx.  The OpEx benefits exist in:

1.  Managing one instance of the application.  (patch/ upgrade it one place and roll it out instantly to everyone)

2.  Reduction in help desk calls. (the ability to reboot the image and get a fresh OS fixes all kids of problems)

3.  Not having to go to each end point to fix issues or load applications.

4.  Products like EdgeSight and AppSpeed take troubleshoot to a new level and reduce the finger pointing and guess work, fixing issues much more quickly with less politics.

5.  Eventually PC have to be replaced even though they will last much longer in a VDI/TS model, Organizations have the ability to use a thin client device which is solid state and 25% of the price of a full desktop.

MS license is going to be the same if you have a traditional PC or not, TS or VETC license will get you either way, You are taking the cost of the PC and moving it into the datacenter, but the cost of managing it is drastically reduced.  Take the time to spend on revenue generating projects rather than on maintenance.  

CapEx being equal and OpEx being hugely a head, it is a no brainer for most companies particularly in an economy where you need to get more done with less.  

Just my $.02


This just underlines the fact that the average person is an idiot who believes all the VDI Hype out there, and does not understand that there are more models than VDI for desktop use cases. MS/Quest/Citrix etc all get this. Citrix certainly has the most models.

VDI costs more, period, cost is not the driver. Brian points out that this customer is going to use persistent images. Why? because at least he/she is smart enough to realize that this power of one is not applicable in their case. 99% of the people I have spoken to start on persistent so they can use their existing tools and process. That means lot's of storage that is complex and $$$ or you use your brain and figure out that local storage is good enough and think about things like iSCSI etc for the future with DASD.

As soon as you put stuff in the data center it's expensive. How much does it cost to host a server in a data center vs. a cheap desktop? You effectively are now using 2 CPU's, one on the client and one on server. F all this 1/16 of a CORE per user vendor fud, no one can ever implement that in the real world. How much does it cost to provide power and cooling in the DC? How much does DC real estate cost? Sure you can outsource to pass on capital costs, but who's doing that day one. What makes one think that they won't need to replace servers in the DC in time either? Sure there is a density rationalization but it's mot significant in a real world deployment.

As others have pointed out VDI is about agility not cost. The other models such as TS represent much lower cost if that is where you are going. If like most people you have to use persistent, you really need to step back and ask why, are just apps or any other method good enough. Perhaps all you need is better PC management, so things like PVS are options. I'll challenge anybody at VMWare, Citrix, Quest, Microsoft to show me really world complex environments at scale where VDI is cheaper from a pure CapEx or OpEx perspective today.

In time of layers ever get's figured out, OpEx may get cheaper. If power of one is what you want for OpEx, go do TS it's cheaper. If some apps don't work you still have options if you understand the technology.

They main problem, Brian I think you are guilty here as well, is that there is just too much focus on a one model. VDI. It's a term invented by VMWare that   they want the world to believe is cheaper. It's not and the conversation need to move beyond VDI. Perhaps let's invent our word for it. Citrix likes to call it Desktop Virtualization, I call it DV for short. I don't want to use a vendor term. I think of it as Centralization and or management of the desktop. CMD. It means you can centralize hardware for agility, and move toward or use better management depending on which model you use. Or you can choose to remain with the distributed model and manage centrally, think client hypervisor, PVS, streamed apps etc. In other words manage the hosted or distributed complexity with centralized management.


AppDetective, your last few sentences summed it up for me.

This is what I see for the future of Desktop Computing that will inevitably happen. A seemless and dynamic approach using server/workstation hardware as one to deliver the best possible user experience the most effecient way.

Minimizing vendor lock-in and proprietary hardware will maximize your available use cases making a much more scalable and future proof solution.

$500 can go a long way to get decent hardware at the end point. If you dynamically use server/workstation hardware as one you would use the end point device as an effective thin client as well as a fat client that can be taken offsite. To me this the athletic client that Shawn Bass wanted, but that's just my opinion.


First of all, I guess it depends on how you define VDI.  I don’t define VDI like VMware does.  I define VDI, Desktop Virtualization, etc as a combination of different technologies focused at first delivering the best user experience and second driving down costs (through operational efficiencies, CapEx efficiencies, etc).  This would includes TS/XenApp, app virtualization, server hosted virtualization, etc.  Whether you are taking Quest, Microsoft, or the Citrix model they generally get the big picture that it has to be a hybrid of all of those types of solutions based on the customer needs/demands.

So how does this affect CapEx? Using the model above large enterprises in industries such as Retail, Banking, Manufacturing, Education, Healthcare, etc can move large percentages of their organizations to the best cost model of them all, TS/XenApp.  From there you can start to look at the end point.  I don’t want to get in the debate of whether or not thin client vs. PCs/Laptops so let’s examine PCs/laptops.  Most large enterprise organizations (10-15k+ devices) that have a well managed traditional desktop infrastructure still need consistent hardware models to avoid image/maintenance nightmares.  Whether you look at Dell, HP, or whomever all of them carry “enterprise class” desktops and laptops.  What is it that you get with these devices? Consist hardware configurations, longer life cycles, stable images, few expected driver updates, etc all things that would turn hardware/OS management into a nightmare and things not available on small business or consumer class/grade endpoints. Desktop virtualization starts to make that a mute point.  Do you really care about hardware standardization now? Add in the futures of client hypervisors and it opens it up even more.  These machines have the same powerful chipsets, graphics processors, etc but they are 30-40% less in price.  You go from an $700-800 desktop to a $400-500 desktop, a $900-$1100 laptop to a $500-700 laptop.  Do the refresh math at the typical 25% every year.  Add in now potentially you do not have to refresh ever 4 years because the device doesn’t have to meet the next operating system or office automation hardware requirements.  It just needs some type of client software to connect to its desktop virtualization infrastructure.

This is just one area where ( especially for large organizations) there is huge potential for savings.  Add in the fact that storage costs are going down, Intel/AMD are finding ways to give you double the processing power with a fractional increase in power and cooling in the data center.  I do believe that 2010-2011 are going to be the years that you will see desktop virtualization show lower TCOs (both CapEx and OpEx).  My 2 cents also.


Firstly a hello to the community who frequent this blog, and thanks to Brian and Gabe for keeping this blog going.

I’ll keep vendor fud out of this and provide my view since this thread struck a cord with me. I agree that there is not one model for Desktop Virtualization, and that is something Citrix are addressing with XenDesktop 4. Getting the world to understand the pragmatic reality of multiple desktop delivery models is something I spend a lot of time talking to customers about. Key questions I always try to understand when I meet with a customer are what they are trying to do and why they think their particular architecture of choice makes sense. I find rather than trying to scream my product is better than theirs etc, a real world conversation about why first and then how usually leads customers to their natural conclusion of what works best for them and removes vendor bias from a discussion. I have found in reality this results in customers wanting multiple architectures to choose from. The other realization that many come to is what happens if they choose the wrong architecture or if their needs change over time? That’s where the flexibility of multiple models with a single license (I know, I know lot’s of opinion and debate on license models) helps protect investments moving forward.

What label you want to give this is up to you. I agree with @appdetective that centralized management for either the hosted or distributed models is key to reducing complexity and I’d add that leads to lower costs in time.

Regarding VDI. I 100% agree that most people are leading with it for agility reasons. Cost advantages are around business flexibility. Some common examples I hear are:

Built in disaster recovery to your desktop model.

Mobility, M & A, Globalization, Data center centralization.

Centralized staffing models.

Investment for future mgmt model

These are just a few. Despite this, it does not cover 100% of the use cases and not all are willing or able to make the investment now. These folks needs can be met by the other models, and it’s very easy for them to understand, once they understand that they have and need choice.

I also agree with @Icelus, that customers want to protect from vendor lock in. I see two reasons for that.

Forces the vendors to keep innovating so customers get the features they want. I.e. Competition is good.

Price leverage. I used to be a customer so totally understand leverage to drive costs down or get more for less :-)

In this respect being able to leverage things like multiple Hypervisors, work across many devices, OSs and networks as possible etc is certainly a key focus area at Citrix and something that I will keep pushing internally. Openness helps avoid lock-in. Before @appdetective chimes in and reminds me that to be open that brokerless connection is a must have. I hear you sir, Citrix is listening. It’s a question of priorities and what a broad set of customers want first. In other words what’s commercial to do, when, why, if etc.


@harrylabana Good to see you posting here and you're right I am going to bring up HDX connect. I certainly believe you get this space, but I don't believe that the XenDesktop management team get's the space. Today Nov 16th Citrix launches XD 4 and the best description I can give to any request that has ever been made of the XD team, speaking to their GM and PM team is a bunch of high fives and no action. That is the only way to describe the GM (Raj Dhingra) and PM (Amitabh Sinha) at Citrix doing XD. Arrogant and stupid, and nothing like the people I have dealt with at Citrix for years. Too much West cockiness vs. real world having done anything in our space IMO. Is Templeton blind to put the future of Citrix with them?  Yeah I called it. I hate to name names, but hear these people speak ask them anything and you will get a high five and no substance. It is their fault that XD is so bad. They don't get it and have shown they don't so many times. Is Citrix really going to let these idiots screw up their product line. I really worry with these two clowns.

That's the danger Citrix faces. Some of these morons in the XD team think they know the space and can predict what customers want. Let me tell these aholes something. Nobody asked for your crappy non scalable broker. It sucks so bad. XA is a much better product including the broker. You are completely missing the boat. Nobody marginally complex is implementing pooled desktops. That means not all want or need to use your broker. Just think if I was a small business service provider and have my own stuff. Why should I rebuild everything just to lock-in to your crappy broker? You need to provide choice and do something with your broker to add value not get in the way of a simple connection. It's obvious you morons.

Reverse seamless has been asked for years, and you have done nothing there either. WTF *** excuse do you have there? So you really believe that there will never be a need to run stuff locally? Do you think that there are multi media standards that will allow you to run things remotely all the time. You will always play catch up. It's so f'ing obvious if you get the space. Again just high fives from these aholes.

Yes you need multiple models, BUT you need flexibility in how they are consumed also. EOM.


Someone get AppDetective some crunchy Wheaties already!

While I'm a little less angry than AppD, I can agree on many of the things that he's saying.  Personally I think HDX Connect and Reverse Seamless (Project Alice) should both be no-brainers.  I can also share frustration with some of the XD product management.  It seems like there's times that some of the XD guys don't get why you'd use VDI vs TS in the first place.  Of course, the XA guys are so committed to the TS model sometimes they don't get why anyone would use VDI.  I can't even imagine what the holiday parties must be like there ;)



Hi AppDetective -

I have been asking a lot of folks @ CTX about revierse seamless in the last couple of weeks as well.  Get a lot of hem and hawing and no real answers.  Still getting people saying that it is not really a priority and that their new Voice Codecs are the answer to everything.  Fact is the voice codecs were VERY easy to improve because they had REALLY SH$TTY codecs to start with.  Still getting the push for Voip via streaming or via published app.  Also, everyone there is on the same message about VDI, which is a big part of their new Licensing Promo's.  Basically its cheaper for customers to renew licensing for XenApp if they purchase XenDesktop (VDI) licenses.  Which means that the company can tell shareholders that the market is picking up on VDI demand, when it is not actually the case.  I see your frustration and I raise you.


I already have VDI implemented and it's saving us CapEX and OpEX.

However, each scenario is different and your milage may vary.  We are a small healthcare company, not a "complex environment at scale" but we are certainly "real world" and I'm sorry if the current 'popular' models don't work for you people.  It seems your all overlooking the small business market in terms of deploying VDI.

The whole point is to find a solution that best fits the particular business requirements and not rely on your favorite vendor because that's all your experienced with.  

The tag-line is  "Your independent source for application and desktop virtualization".  Maybe I'm wrong, but it seems most of the commentators are not actually end-users but solution providers, industry pundits, or independent consultants...funny how that works.

And I don't give a darn about the acronym. VDI works just fine for me regardless of who invented it.

Rodd Ahrenstorff


Regarding Reverse Seamless, isn't this what the whole XCI or Xen Client is all about.. Now, exactly how and in what way this relates to VDI or XD I don't know. I think one use case is to mixture of technologies. Time will tell.

XCI/XC is not the only kid on the block. VMware is also developing their Client Hypervisor, and people here allready know about Neocleus, Virtual Computer and what else is there cooking? A curious mind wonders what's in the stew over at Redmond


My company has tried, and are still trying, to sell VDI and Citrix XenDesktop specifically to a whole bunch customers with 500-20 000 desktop fleets.  CapEx or OpEx doesnt matter as long as they can reduce their desktop cost. In 98% of all cases they fail to sell anything, but traditional thick desktop mgmt. Why?

Based on my personal experience with 4 of these customers.

Customers of 2009 with thick PCs have well managed desktop environments using a huge MS toolsetbox for desktop mgmt + well known industry best practices for desktop mgmt, which most enterprises with >500 desktops do anyway. Hence the OpEx story for VDI isn't relevant, or if it is, it is just marginal and W7 improvements muddys the water.

Customers have huge storage costs, having spent $$$ on SANs for the last 5 years. They try to scale back on storage usage. VDI doesnt fit there as they understand they will have to give users persistent desktops.  

CTX+MS licensing story and previous presentation protocol issues for SBC have scared them.

MS have given customers new solutions like DirectAccess, which enables any client to be on at anytime as long as there is an Internet available, hence OpEx are bound to be reduced for traditional desktops even further (add Desired Configuration Management, MDT 2010, ++).

Customers considers VDI at peak of inflated expectations and would like things to mature a bit before they jump on board.

Yes, customers are pragmatic so there are still virtual desktops deployed. They tend to end up for minor use cases like SW developers in India or non-employees requiring a corp desktop (<1-2% of total volume).  

I havent met any customer who wants to discuss VDI, for the major population of desktops, for any other reason than cost cutting.

Sounds like a good idea to sell XenDesktop for "free" if you buy XenApp.



Regardless of your technical point, naming people the way you have here is simply not cool.

IMHO, pull your head in.


I agree with Rodd.

When a company is already investing in data center, network, and desktops, it's actually cost effective to migrate a large portion of desktops in to the data center.  

AppDetective, I think your logic is slightly misleading because when you ask how much it costs to cool a data center you seem to be assuming that someone is entering the data center to do VDI and VDI only....

On the contrary, most business are looking at their infrastructure and asking 2 primary questions: 1) how can I alleviate the costs associated with disparate machines across multiple offices/sites?  2) How can I better utilize the expenses being committed to the data center?  It could even be said that IT management is looking to be more effective at the support desk level too because let's face it, it's not as easy to deploy and support desktops as you claim....

If you consider that VDI does seriously address these pain points then it makes sense why so much attention is being given to it.

Regarding Citrix management team, I wouldn't have the first clue.  I can tell you this though, I don't see XenDesktop as a contender when put side by side with VMWare Infrastructure....

And a final note about capex and virtualization.  I have to agree with you, that generally, virtualization done *right* is quite expensive,  On the contrary though, a small business able to invest in 2 decent servers, 2 NAS appliances (one for backup off site) can run ESXi quite cheaply....and VDI is not that expensive assuming you are fine doing persistent desktops without link cloning....


But John, you could deliver a centralised environment that could be delivered to desktops or remotely, without 2 nas appliances, without managing esx, without provisioning services without managing each individual instance.

You can do it by installing terminal services on two servers. Maybe you just use terminal services, maybe you add XenApp - or not - Ericom, Propalms, Quest  all have solutions as well - and all of that together would be less time, effort and resource to manage than an equivalent VDI environment. Its not even contrary, its fact.


@John. People want to reduce DC footprint not increase it. Are you kidding... That's what the VMWare sales team will have you believe, it's BS. To answer your points 1) I agree centralization is one of them, but there is a cost to that, that is > than distributed today unless you are VERY simple, or go for massive density which has it's downside as well. 2) Sure, so why not use TS which is even more efficient? Your  VM Infra comment is just wrong sorry not personal. That Infra manages a single Hypervisor, Citrix/Quest and MS can run on top of it. So that's BS. Are you telling me View scales? It's as bad a XD with a crappy display protocol that no WAN accelerator can consume either. I am sure all the partners who sell riverbed are real happy with PCoIP.

@Rodd. Great if VDI works for you, nothing wrong with that. It would be helpful since you are a simple environment to help us understand why you wouldn't use TS. App Compat is a BS answer BTW. There are good reasons.

@Clayton. I debated that, but something had to be said, since they are such roadblock to progress for so long. I have no issues pointing out the good guys at Citrix either and I do. Also to Brian's tweeters who think I should reveal myself or us? Well then I/us could not do what I/we do. My/Our intention is to say what we feel and not hold back. That offends sometimes. Some people hate it, and that's fine by me/us.

@Grolpels. you are wrong. I know tons of people implementing for the masses the VDI model NOT for cost cutting. They want to connect to their sessions where ever they are. In fact I know of nobody at scale implementing VDI (hosted Desktop OS) FOR cost cutting.

@Glenda @Kimmo Reverse Seamless is about when I want to host and run some things locally with or without XC and have a user experience that is seamless. It solves for many use cases. It needs to become a feature to allow for flexible use of some local apps with the majority remote. There is just no way that all things will run well remote, and there is no way that running all locally is perfect either. It's a blend, and understanding that for many use cases is key and the smart thing to do. Something the XD leadership is not doing and clearly don't get how the real desktop world works.

@Glenda on your point about VDI motion at Citrix. I think in their new XD 4 model the real data point to watch is what model are people implementing at scale and why? It's very easy for Citrix to say XD 4 is huge, but if they are using XA which is part of XD 4, then it shows VDI is not the model for many. That granular data I am sure they will never share, but WE the community should help get a sense of that. Not that it matters frankly if XD 4 is a success with multiple models, because I think it is the right strategy. My gripe is with VDI  false expectations and outright lies regarding TCO of the pure VDI model for all use cases when the reality is very different. After all this thread is about TCO of VDI.


@John, my guess is that you sell VMware and/or only use VMware and have no experience with Citrix.  XenServer might not be able to hold up to vSphere in terms of features, but XenDesktop kicks View in almost every category.  Hypervisor is almost irrelevant unless you count overcommit.  Gotta agree with AppDec on that one.

@AppDec Gotta agree with you on the VDI vs TS model.  I think the move in XD 4 by Citrix is mainly a marketing one.  The more customers start talking about the catch phrase "VDI", the more customers are being lead back to TS/XA for the reasons you mentioned.  It is the best model for 90% of end users.  All Citrix did was make it easier from a marketing perspective to make the leap by putting XA within XD 4.  


AppD you rock man! I responded to some of your comments on my VDI WTF blog.. but take it easy on me we are on the same team here!! Plus I am the Sitg and you stole my picture, but I'll let you borrow it.......

I agree with almost everything Shawn and AppD have posted on this topic, but I feel VDI is useless to begin with.


I don't sell VMWare, I am an IT Director for a corporate end-user!

We use Citrix.  We have used Citrix longer than VMWare actually.  I used to be far more active in the Citrix community than I have been the past couple years, but I assure you I am a big fan of SOME of Citrix's products.  Mostly, Presentation Server.

Guys, ever try integrating XenDesktop in with ESX or vSphere?!? ROFL!  So I will give you the argument that XenDesktop might be better than View assuming XenDesktop is running on top of XenServer (and that is ONLY because of ICA).  And....c'mon, don't even try to compare XenServer with vSphere!!!!!  

I will agree from a protocol perspective ICA is better.

@AppDetective: if you are saying TS scales, then by sheer logic View scales being based on RDP.  I can't say regarding PCoIP though.  BUT, View and other VMWare products are highly dependent upon the storage infrastructure they run on.  YES, I argue that View scales when deployed alongside a DECENT storage solution (I am thinking NetApp or 3PAR, with cache acceleration, etc.)....because the problem with View is not the architecture, but the storage, network, and I/O.

Listen, this is why Cisco jumped the market by releasing UCS and their 10Gb DCE standards with a partnership between VMWare and EMC.  Were ANY of you at the Cisco seminars at VMWORLD this year?  Completely mind blowing benchmarks.  What they do for servers is so EASILY crossed over in to the VDI scenerio.  Everyone knows the pain points in virtualization and SCALE is only so because of storage and network.  Well, storage vendors are stepping up big time.  And the layer 2 and layer 3 components are coming full steam.

Now, regarding your comment about shrinking the data center space I would point out that you are only HALF right.  How can you agree with yourself when the entire industry is moving toward "cloud computing".  I know that term is loaded, but remember when this was all based on disaster recovery?  The entire point of moving to the data center was to increase MOBILITY in the work force, to lower other associated costs, and to better leverage the data center investment we HAVE to make.  SO.....the reason we are trying to SHRINK the footprint in the data center is two fold

1. Eco reasons (green)

2. We have we are relying more upon centralized computing models it only makes sense that we are looking for thinner, faster, and lighter systems....TO MAKE ROOM FOR MORE SYSTEMS.....


We saw a 70% recovery in storage accross our VMWare datastores when we deduplicated it.

Anyway, I respect your opinions.  This is what I love Brian's site and community for.  Generally, I have always found such incredible professionals with intelligent opinions here.



Just a couple of points John.

" What they do for servers is so EASILY crossed over in to the VDI scenerio."

Not necessarily. The storage and network requirements of a desktop OS workload are, or can be, very different from a typical Windows Server 2003/08 workload and must be considered up front. I would agree that they can "cross over" but it shouldn't be approached with a preconceived idea of being "easy".

"when the entire industry is moving toward "cloud computing""

Hmmm, not sure yet.

Cheers, Clayton



If you've read any of AppD's posts, you'd understand that he's not a fan of XenDesktop because of anything that it is other than the ICA protocol (this is part of the reason why one of my customers bought Standard Edition of XD).  AppD (and I) are not fans of the broker, nor the pool manager, nor the Provisioning Services as part of a VDI implementation.  It's not that any of those things are wrong from an approach perspective.  It's just that they're immature at this time.  The same goes for View.  Although from the View perspective you've got RDP (which makes it a loser to XenDesktop) and more recently PCoIP (which I'll withhold judgement on until I've seen the specific results myself).

Oh and don't assume XenDesktop runs ontop of XenServer.  Many people run it on ESX because that's the established Hypervisor.  ICA is really the reason most people have bought into the concept of XenDesktop.  Of course, that's just going to irritate me that again that Standard Edition is gone and I now have to license "VDI-Edition" at 2x the cost.  That's a rant for another day I guess.



@John - Also, I'm not clear what you meant by "if you are saying TS scales, then by sheer logic View scales being based on RDP."  Can you elaborate?

I think you might be assuming that AppD says TS scales because it uses RDP and therefore View scales because it uses RDP?  But RDP is just the display remoting protocol.  It has nothing to do with the scale.  For a scale conversation you're still talking VDI vs TS and TS will almost always win those scale conversations.  VDI beating out TS will *probably* never be about scale, it'll always be about maximum app compat, vendor support, personalization, potential reduced complexity, some user/exec's emotional attachment to a full PC OS, etc.



I was going to ask @John to clarify the RDP scaling thing too.... Shawn beat me to the punch.

IMHO VDI in general has more scaling issues than SBC due to running completely isolated multiple instances of the base OS and not having got that single image thing really right. Although I have not explored the new release of View yet...


@john. Hello. Just to add to other huh questions I like to address your comments about cloud. I'd respectfully disagree. nobody is rushing to the could. What people want IMO is more efficiency in the data center. It starts there with Server Virtualization consolidation etc internally first and forces you to rethink how to manage this new reality. I know my organization sucks at this, and it will take time to figure out how to manage server virtualization at scale. In fact I spoke to somebody recently who found it was cheaper to just manage out excess server from their environment and improve process than to jump over virtualization. This person achieved 30% reduction in footprint since they had so much crap built up over the years. Adding more servers to go to virtualization when there was already a sunk cost made no sense.  Cloud as soon as you go public is more about consuming utility services that you don't have to have sunk cost in. There is no way even the large enterprise customers are going to get the scale economics of the large cloud providers. Spending power will shift over time to these service providers IMO and that will be cheapest place to buy compute power.

Also XenServer vs. VSphere. Well I read the Burton group report the other week showing XS as enterprise ready, I understand XS does not have memory overcommit. I've tested both, run XA and XD on ESX today which works fine minus the crappy support from VMWare. I also find for Citrix workloads XS is fine. XS runs better on older hardware, something that the virtual reality check guys did and I picked up on and can also confirm. Good work there. XenServer is now free and there is a bunch of management there as an extra. So not only is the hypervisor a commodity but so are many of the management tools. Hypervisor diversity especially if you believe in the future of the cloud and have seen some of the Azure announcements today is an emerging reality. VMWare LOCKS you into one way. They are heading the way of Novell and Netscape. Open and standards base will be key for the cloud. Don't get me wrong, ESX is a great product, but the price point vs. what the others are building and decreasing lead, lack of open management tools etc just smells like a train wreck.

Regarding scaling. The model of VDI being touted is too complex with shared storage and the $$$ are not there. Better to use local. That's what XA does and it's fine for many but not all use cases.

@Eric I feel you are wrong when you say VDI is useless. It has saved my butt many a time and many of my friends. We would have been screwed on TS for one reason or another. Sorry dude, I think you are missing a massive use case that meets the needs of some. It's more than niche as you elude to in your article, there is real value to drive business use cases forward. I do however still think that XA is good enough and cheaper for the vast majority. I think @harrylabana makes a good point above in his post that people can pick the wrong model and have changing needs, so a single license to handle that is good, I agree. However Shawn I hear your XD standard rant, I just think that is a small percentage and enough to get be all bothered like I did last time. Anyway Eric,  good to see you having an opinion and getting people to THINK. That's the problem too many sheep and not enough thinking and debate, disagreement to help us all learn from each other. I'm done with this thread :-)



I wish sarcasm came off better in text, or I wish I was better at conveying it.

Truth be told I do not think VDI is useless, I just think the use cases being floated around by the manufactures are far reaching and typically better served in a TS/Winframe/Metaframe/PresetationServer/XenApp environment.

I don't agree with the thought process around VDI for large scale deployment typically.

Thanks for the Kudos too many organizations CIO's and CTO's are just going along with the marketing or whatever they are being sold....Some pretty smart people in this space but sometimes I think they take the path of least resistance.

To the article's point if VDI is finally cheaper for CapEx where does that leave SBC with greater user densities....