VMware announces View 4 with software PC-over-IP and vSphere 4 support

Listen to this podcast

Today VMware announced the most recent version of their desktop product, View 4. This new version has two main new features: A software-based, PC-over-IP remoting protocol is built-in View now runs on vSphere 4.

Today VMware announced the most recent version of their desktop product, View 4. This new version has two main new features:

  • A software-based, PC-over-IP remoting protocol is built-in
  • View now runs on vSphere 4. (That’s right—it didn’t before!!)

The remaining major features of View have not been updated, including:

  • View Composer is the same as before.
  • Offline VDI is still experimental.
  • ThinApp is the same.
  • VMware CVP (the client hypervisor) is still not available. (1H10 is the current ETA.)
  • Windows 7 is not supported. (It’s there as “experimental.”)
  • RTO Software’s Virtual Profiles product (which VMware announced they would OEM), is NOT yet included.

While there are many advantages to basing your VDI infrastructure on vSphere 4 (which we’ll look at later in this article), it looks like the only major new View-specific feature is PC-over-IP (despite the fact there are hundreds of people on the desktop team).

VMware’s software PC-over-IP protocol better be pretty good!

How good is the software PC-over-IP?

I have no idea. I haven’t used it yet. I did just get the release code from VMware for View 4 over the weekend, so Gabe and I will put it through its tests and see if PC-over-IP is as good as VMware claims it is. For the record, I don’t think PC-over-IP has to be as good as HDX—it just has to be good enough for people to use.

And certainly VMware claims it’s ready. Everyone there seems really proud of the work they did getting PC-over-IP to run in software. They claim their implementation works on the LAN and the WAN and across wide samples of latency, bandwidth, and packet loss. They talk about how it’s dynamically adaptive, both changing as network characteristics change and based on what’s on the screen (apps, text, Flash, etc.). VMware also points out that PC-over-IP remotes the entire client experience, so you get stuff like multiple monitors and USB redirection.

The idea for View 4 is that PC-over-IP will replace RDP with TCX as the “go to” protocol for View and will be used by everyone. My understanding is right now there’s only a client for Windows, but they want to release a Mac client in the future. And software and hardware PC-over-IP clients and hosts will be interchangeable, so you could also use a thin client device with a hardware chip in it with the software PC-over-IP of View 4.

In fact VMware is really proud of the ecosystem developing around PC-over-IP, even releasing a press release talking about it. (Although for some reason the first partner mentioned in the PC-over-IP ecosystem press release is Cisco, and after talking about their WAN optimization which has nothing to do with PC-over-IP, they talk about the security of the Nexus 1000V!?! Come on.. this is a press release just about PC-over-IP!) But if you get through that weird Cisco part of the release, you’ll see HP, Dell, Devon IT, and Wyse talking about how they support the protocol.

Obviously View 4 will still support RDP-based connections as well for older or non-Windows clients. VMware had previously some of Wyse’s TCX extensions for RDP to help make it a better experience. I assume that agreement continues, although I don’t know for sure. (I asked on Friday but haven’t yet heard back on that. I’ll update this when I do.)

Running desktops on vSphere 4

VMware will be spending a lot of time talking about how good vSphere 4 is and why it’s the best platform for desktop virtualization.

That may be true, but what’s kind of weird is that other desktop virtualization products like Citrix XenDesktop and Quest vWorkspace also have the ability to run on vSphere. So a “win” for vSphere, while certainly a “win” for VMware, is not necessarily a “win” for View. (Does that make sense?)

I guess really the win for View when it comes to vSphere is that when you buy View, you get the highest-end edition (Enterprise Plus) included with your purchase. (Brilliant move, btw!) Customers who chose vSphere for with non-View VDI products would have to buy vSphere licenses on top of whatever desktop product they’re using. (And vSphere isn’t cheap, starting a $800 per processor and climbing to a whopping $3500 per processor for the Enterprise Plus edition that comes with View!)

And even though there’s nothing stopping customers from buying the $3500 per processor license for their VDI environments, it’s most likely that customers will opt for one of the cheaper editions, believing that most of the high-end virtualization infrastructure features included in the top package don’t really matter for their VDI desktops. So while that might be true, it also means that since both VMware View editions do include the highest end vSphere edition, you really need to look at those features and consider them as part of the value you get with View that you wouldn’t get somewhere else. For example, VMware Distributed Resource Schedule (DRS) watches server load and dynamically moves VMs (via live migration) across physical hosts. This could allow you to consolidate VMs to fewer servers to shut down hosts to save power. And since DRS is part of the super expensive version of vSphere that’s included with View, you could claim that as a feature of View even though it’s technically made possible by the underlying vSphere.

Performance improvements of desktop VMs on vSphere 4.

Of course VMware believes that vSphere 4 is the best platform on the planet for running any type of virtual workload, but they’ve also made many changes and improvements that specifically help desktops. And even though vSphere 4 isn’t new, the fact that View 3 wasn’t supported on it means that the desktop-specific enhancements have gone unnoticed so far.

One of the big thrusts of vSphere 4 was a focus on making improvements for targeted workloads (like desktops). To that end there are more than 80 specific improvements to vSphere 4 which ultimately lead to a doubling of the number of desktop VMs per core they now recommend to 16! They’re also talking about major improvements in the number of VMs per LUN and a drastic reduction of VM boot time and mitigation of boot storms.

Does using View on vSphere give you anything that using another VDI product on vSphere wouldn’t?

It’s clear that VMware will be talking a lot about vSphere when they talk about their desktop offerings. What’s unclear is whether this focus is because vSphere 4 is really so much better than everything else out there or whether it’s just because VMware doesn’t really have too much else on top of vSphere compared to the other companies whose desktop products also run on vSphere.

Notwithstanding the pricing advantages of getting vSphere through View that we previously discussed, I specifically asked VMware whether there were specific advantages customers got using View on vSphere that weren’t available in other companies’ products running on vSphere. The answer was that while everything View does vSphere is done via public APIs, VMware is the only company at this point which makes use of all the APIs. In other words, View is better because it’s made by the people who also make the underlying infrastructure. Thin provisioning is an example of this. Thin provisioning is now built-in to vSphere 4, and the only way to get vSphere’s thin provisioning capabilities for desktops is to buy View. (Although to be fair, several other desktop virtualization vendors have their own thin provisioning products which can be used with vSphere 4.)

A few other small changes

Since the only real new features of View 4 are PC-over-IP and vSphere 4 support, I asked VMware for any other changes or improvements that were not due to improvements in the underlying vSphere infrastructure.

One new feature that was mentioned was the addition of proximity-based security policies which allow different security configurations depending on where the user is connecting from. Unfortunately my understanding of this is that it works by allowing different connection servers to have different policies, so the idea is that when a user is inside the firewall they connect via one server, and when outside the firewall they connect via another. So I guess this is cool, but not as simple as just applying policies based on the client network. You need to build multiple connection servers to do this.

Another new feature is single-sign on improvements, where the user can just sign in once to the desktop View client or the web portal and their credentials are automatically passed back to the VM.

View 4 Weaknesses

I think there are two main weaknesses with View 4.

The first is that it’s VDI-only. In other words, you can only use it for remote server-hosted desktops running in remote VMs. You can’t use View to put desktops on physical machines. There’s no supported offline solution. (The experimental solution that’s part of the product is total crap and not supported.) And you can’t use View to connect to Terminal Server sessions or blades. (Ok, technically View can broker connections to those types of hosts, but they have inferior capabilities, such as no support for software PC-over-IP, ThinPrint, or the Wyse TCX extensions.) So View is truly a “VDI only” product while companies like Citrix, Microsoft, and Quest are trying to build products that encapsulate all your desktop use cases. (Of course that means that View is a simpler product too, which certainly has its virtues.)

The other big drawback is the vendor lock-in. With just about every other desktop delivery product on the market today, you can choose (or mix-and-match) your underlying virtualization platform and infrastructure. With View, you’re locked to vSphere. This means if you ever want to change your underlying virtualization platform, you’d also have to throw away your desktop infrastructure too.


Moving forward, VMware is promising that View and vSphere will be in lockstep, and that we won’t have the situation that we had the past few months where View wasn’t supported on the latest platform. VMware is also starting to talk about the improvements you’ll see in view, like mobility, full full deployment of offline, Mac support, and CVP.

View 4 will be available on November 19 in two editions, a lower-end one for $150 per CCU and a higher-end one for $250 per CCU. (This is the same price as before.) I asked VMware about how the CCU licenses work with offline, but I haven’t heard back from them on that yet. I guess that’s why that feature is “experimental!” :)

But they do claim that the acquisition costs of desktops with View is now cheaper than it ever has been, and in fact can now be cheaper than physical desktops. (This is based on a combination of View pricing plus the performance improvements in the vSphere 4 platform.) Will it work out? Time will tell. I haven’t had any hands-on time with the product yet, but rest assured Gabe and I will be banging on it all week and we’ll share what we learn as we learn it.

What about you? Have you used View 4? Have you run desktops on vSphere 4? What are your thoughts?

[TECH NOTE: I just uploaded the audio version of this article, if you'd like to listen to it instead of read it. Check the "attachment" up at the top near my avatar.]

Join the conversation


Send me notifications when other members comment.

Please create a username to comment.

Good write up Brian. It's amazing to me still why anybody would look at such a crap product with a bad strategy. Brian hits the nail on the head by pointing out this is a VDI only solution. There has been plenty of debate and people who really understand this industry know that VDI cost are greater. The single image concept is just not there yet. So simple VDI will cost more than your desktops. Granted if your drivers are not cost, agility etc then VDI can make sense. I've said it before and so have others, XA/RDS is cheapest VDI solution.

PCoIP. So much fud out there. From what I have seen, it sucks bandwidth like no tomorrow, looks good on the LAN though. The other thing is all this Wyse crap they are fooling people with. VMWare expects people to get locked into crappy thin clients that are $$$ to deliver their WAN story. Even with TCX the BW is high on PCoIP just like RDP. Does VMWare expect all the morons who use View to switch from RDP to Wyse thin clients to deliver the WAN. It's UNWYSE to use loud mouth CEO thin clients and get locked in. Add to that I believe that MS will not support Callista on ESX and it's even worse.

The rah rah about them adding VSphere is also crap. Not only are VMWare locking people into an ESX only desktop solution they are trying to lock people into custom management framework. Guess what MS CAN'T be ignored and I'm even considering some XenServer in my shop. I also have physical blades that can't run a Hypervisor and I am not going to the through them away just yet. That means you need a management layer to manage all. Add to that the sucky support VMWare gives everytime a MS issue occurs and it's just horrible.

So the good news is that is competition and good for our industry. But really so much fud with VMWare custom hardware, custom management, and of course all they f'ers require a broker which is not needed and represents lock in as well.


Hmm, although You usually write good reviews, this is more resemblant of a veeeery long, snide remark.

A little more meat on the bones would go a long way.


I'm surprised that View is lagging behind.  VDI is more than just a VM desktop.  Not saying XenDesktop is the best thing ever, but definitely several steps ahead.

Curious how well this fits into BYOC projects companies are implementing?



@dkvello, were you talking to me, or AppDetective? Because I want to be clear that what I wrote is NOT a review. I have not used the product yet. This article is based this on some interviews I had last week with the product folks. A full review will come in the next week or so.

And to be honest, I'm actually excited about View 4! All the vSphere stuff seems really cool, and VMware definitely deserves credit for that. But as for the rest of what I wrote, I mean come on.. the only new feature apart from vSphere support is PC-over-IP? Jeez.. Where are all these other things they've been talking about for over a year now? (CVP, Offline, etc, etc...)


This is great news. PCoIP is set to change the VDI landscape. Citrix and others wil have an extremely tough time trying to match what VIEW offers - REAL desktop experience (i.e. quality, multimedia, local devices, USB, etc) as opposed to legacy ICA protocol that is not a drect desktop experience delivery (i.e. it goes through a "thin" protocol and is thus lacking the same level of "directness". ICA is great for dialing in t use Word. Legacy protocol eclipsed as we speak.


This is fantastic.  Thanks for your hard efforts Brian!


Commence VMware / Citrix fanboy debate.  I'll opt out of the slander and we'll wait and see real results soon enough.



It would be very nice for a side-by-side comparison of XD4, View4 on VSpehere with Windows 7 client loads, or even Vista or what ever is the latest Windows OS VMware actually supports.

I would've have Microsoft VDI in this comparisson, but am not quite certain beein that Microsoft VDI is Hyper-V ony whereas View and XDE at least CAN share a common HV. Maybe in a some other comparisson.

There's a lot more I'd like on the table, but none of them are nearly real at the moment.


Thanks for the write up Brain.

Before I go into my thoughts / opinions, I would like to answer your questions. Yes & Yes.

Caveat…*I know a lot of this can also be done with XenD and ICA/HDX*

1) Software PCoIP :

a. LAN User Experience: The UX on the LAN is great. It truly does what it says it can. I was able to watch a video playing in windows media player and a flash youtube video on one monitor and scroll through a heavy graphical PDF without missing a beat on the second monitor.

i. VM = XP SP3 1GB RAM 1vCPU

ii. VM / Host / Endpoint Impact: There is no free lunch. You have to be smart about what you plan to do with View 4 and PCoIP. Throwing it on older hardware will not give you the most amazing of results. My experiences with ESX 3.5 U3 and older hardware were not as good as they were with ESX 4.0 U1 and new hardware. (I know, hardly a surprise there) Anyway, there is some vCPU / Host / Endpoint CPU impact when running a lot of graphically intensive workloads, as there should be,(this is software, you are going to have to pay the toll somewhere). However, vSphere more than adequately handled the load with no visual impact to the user.  I am anxious to see the host impact when you scale it to 50+ users on a host (only at 20 right now) (Where is Project VRC when you need them :-) ).

b. Multi-Monitor Support: I found the multi-monitor support to be awesome with View 4 / PCoIP. It just worked. The logon box was primary monitor justified, windows expanded to the correct screen – all the time, there was no additional software required like TCX or Splitview. It was actually easier than native windows tools that don’t always remember the settings anyway. I am looking forward to seeing how this will work with Dual Monitor PCoIP TCs.

c. WAN Impact / User Experience: For me, ICA is the benchmark when it comes to WAN and using a remote display protocol. Therefore during my usage of the product I was continually comparing it to ICA, which might not be %100 fair, ICA is still more mature. Based on my experience with what I have done with ICA and WAN solutions, there was always a compromise when it came to lossy / lossless compression and what you really wanted to deliver. I found that with PCoIP I was able to experience an environment that was graphically better than RDP and not as good with no tweaks as a highly optimized ICA Session. I know that is not the most technical of explanations however I am still testing the WAN scenarios and will report more when I have the information. Also, I did not scale it at all. 4 Users is hardly an encompassing test and having to have a VPN connection into my test environment limited the amount of folks I could throw this at.

2) Oinions on the vSphere ESX / View dependency.

This goes back to the “Does the Hypervisor really matter in VDI” debate. As far as View is concerned, let’s face it, View is VMware’s Desktop Virtualization product. VMware is a virtualization company so it would only make sense that at the core of its Desktop Virtualization offering would be its flagship virtualization product, ESX.  Like Brian stated, you get a lot when you go with View from the Hypervisor perspective. I don’t know why anyone would spend that much $$$ on ESX then to only have to pay for another license for the front end. I can understand why View isn’t Hypervisor agnostic and why XenD and others are. Let’s be honest here, Citrix has spent a lot of time delivering desktop oriented solutions vs. Hypervisors and visa versa for VMware. (It is a lot like where we were 10 years ago dealing with WTSUPRN.INF and driver hell just to get printing to work in TS/Metaframe) I feel the View feature set still has some growing pains to go through on its way to maturity. It will get there. I also feel that out of the box full Win 7 support should have been there in View 4 or they should have just called it View 3.5. I really hope they get Win 7 supported sooner rather than later.


While I can not see this changing the landscape as system.fracture puts it (ICA has been delivering all that for years and is just about to improve with their HDX implementation), it is good for VMware to actually convert something from vaporware and actually deliver it.

From all reports it seems to be a good implementation of the PCoIP protocol.


Yay! I get another vendor lock-in with PCoIP (Teradici) if I choose to offload the workload onto hardware.

HDX 3D will give me choices to use non-proprietary hardware for both server and client side operations if I choose to do so.

I like the method of using my existing GPU's if needed thank you.

Technology wise I like VMware, I just don't like their monopoly tactics, it's very narrow sighted. It's almost as if they're run by ex-MS employees, oh wait, they are.


@Icelus, I don't understand your comment about PCoIP locking you into hardware. You understand that the View 4 is a software-based implementation of this, right? So no custom hardware on the host or the client. (But if you happen to have that custom hardware, then it will work.)


@ Icelus and Brian - All your major thin client vendors are also releasing PCoIP chips so you wont be locked into a vendor if you do not want to be.  HP have this on their roadmap :)


@ Brian, I think Icelus is refering to the need to have a Tera1 host card when brokring to blade pc's and rack workstations. In the case where you need a deidicated GPU for workstation replacement and 3D support.  

That is true, hardware assistance or dedicated brokering 1:1 to a blade pc today does require a tera1 host card in the blade.

Personally, I do not see this as lock in really. HDX 3D requires a CUDA based GPU, FX 3x00 or higher. So Nvidia is the only option or choice you have. I also think a CUDA based card cost more than a Tera1 host card  + Nvidia FX 1x000 which covers the vast majority of the workstation market. You obviously have GPU choice with a tera1 host card. I do udnerstand some see it that way though.


@warren. You make a good point. But the real point is will commodity PC vendors provide machines with the PCoIP Tera 1 host card as standard? I don't see that happening anytime soon. Therefore one is limited to a subset of machines, or locked in to UNWyse or something like that. I do agree with you regarding the 3D use case assuming what you say is true (I haven't dug into it), but  for me not a requirement in my environment so I'm looking at the need to use UNWyse for WAN acceleration.


@ appdetective I am not sure I really follow you. There is no dependency on WYSE. Although WYSE is supporting PCoIP, has announced the addition of a PCoIP zero client to their line up and also will have support for the soft client on their devices. The same is also true for all of the other thin client partners. Sorry if I misunderstood.


@ Warren.  What if we do not want to go thin client.  What if I wanted to go a thick client on the desk and have a cut down image (like a kiosk image), would I be able to get a PCoIP card for that?

A question worth asking is how many people are serious in deploying a thin client in conjunction with any of the VDI technologies?


:-) Sorry about the mistake

The one thing that is most dissapointing about PCoIP right now is that there will be no ThinPrint :-(

View 4.0.1 maybe ?


Regarding View Clients with PCoIP support:

Warren Ponder Sr. Product Manager Remote Display Experience had this to say about SW PCoIP back in February:

"Q. Will Linux be supported by the protocol?

A. Linux clients will also be supported in PCoIP same as win32 and win64. Multimedia redirection will also be part of the solution, as is bidirectional USB support."


For more recent information, your contacts at VMware should be able to give you updated information about supported endpoints for SW PCoIP. As the product is not available yet, further information is still under NDA I guess.



"Wyse will also support PCoIP on its line of Linux and segment-leading Wyse ThinOS clients in the near future."


You won't see PCoIP in non-OEM versions of the View 4 Linux client though!




Your facts about HDX 3D are incorrect.  Citrix had originally stated that you'd need a CUDA enabled adapter with 96 or more cores for HDX 3D (i..e fairly high end NVidia Quadro FX cards). However, that's only to offload all of the HDX 3D processing completely into GPU processing.  The reality is that you can perform many of the operations in software (CPU) and use an off the shelf GPU.  I've tested several NVidia GeForce cards as well as ATi Radeon GPUs.  The big question of course is which adapters would be available in particular workstation blades, but I wouldn't put it past customers to just stand up workstation PC Farms in their data center.  I've seen it done many times often for hundreds of units.



Oh and a $50 GPU can get you anywhere from 60 fps up to several hundred fps (even over the WAN)



Ok. Final comment on SW PCoIP on Linux:

"Software endpoints will use the

PCoIP functionality in the VMware View Client for Windows and Linux"


Go figure.


Hi Shawn, Thanks for the correction. That is my bad, I had not read the most recent requirements doc closely enough and missed the mode of operation doing CPU compressions etc.


PCoIP is now software based which is great...

Question is : will it be better than PCoIP hardware based in term of performance ? I hardly doubt it (as dedicated hardware based solution normally do it better)... anyway,

PCoIP/HW should already been benchmarked against ICA/HDX ??? this could give an expectation about what PCoIP/SW shoud be... more or less...



I understand that this is a sw implementation, and it will never be good as the hw though. All of the vendors know this.

As stated earlier by Shawn Bass, there are two scenarios for offloading the workload onto hardware:

CASE I: The HDX 3D for ProGfx server has a NVIDIA CUDA enabled GPU with at least 96 cores and the NVIDIA CUDA 2.1 (or higher) display drivers installed. In this architecture, the GPU codec can be leveraged to get the optimum performance in both LAN and WAN. You can also use the CPU codec with this configuration.

CASE II: The HDX 3D for ProGfx server has a NVIDIA CUDA enabled GPU with less than 96 cores and the Nvidia CUDA 2.1 (or higher) display drivers installed, or does not contain a CUDA-enabled Nvida GPU at all. In this architecture, only the CPU codec can be leveraged to get the optimum performance in both LAN and WAN.


I have no doubt that both Citrix and VMware technologies are solid comparisions, but in the sense of wide spread adoption in the next few years client hypervisors on VT-x will make it hard to replace those currently existing fat clients with  expensive Blade PC's and expensive thin clients. After all, I won't be able to take a thin client offline and bring it home could I?

I know this post will generate a sht storm with AppDetective on how client hypervisors won't hit mass as soon as we all think, but we will have to wait and see on that one.

This is just myself thinking out loud about vaporware though...


Don't get me wrong, thin clients will always have their place, just like RDS/XA.

The more options that are available that meet the business need, the more scalable the solution is.


Am I the only one out there looking for the "Athletic" client?  Don't like fat, don't like thin.  I want something in between.


@ Shawn,

I hear ya, I would even settle for a "Weekend Jogger" client with a mediocre personality. Throw in all that and keep in under 300 bones and we have something!


From the Teradici FAQ (www.teradici.com/.../pcoip-faqs.php)

The differences between hardware and software display compression can be characterized by differences in server CPU loading, network loading and user experience (resolution, display frame rates etc.). Software display compression can optimize for any one aspect (CPU, network or user experience) at the expense of the other two.


@ Shawn @ Shanetech - That would solve allot of issues.  Give me a SFF computer, low power (Atom processor), small SSD HDD, and I will make my own "Kiosk" based image that will work with MY management tools.

I did just google this and what about a 'nettop'.  www.pcworld.com/.../lenovo_puts_atom_chip_in_fanless_desktop.html

That sounds like an 'athletic' client?


My company is nearing the end of a very..very..long drawn out VDI POC that has turned from a VDI project to a VDI\app delivery project for an environment of 25,000 users.

I have been using both VMware’s vdi solution and quest's vdi/app solution extensively and have come to the conclusion that Quest is a great all in one solution but may lack a little on the display protocol. In my opinion Quest with PCoIP would be an amazing marriage... and it is possible- from what I understand is that VMware does not own exclusive rights to PCoIP...they are just the first to deliver. My onsite xenapp admins have verified that using quest to deliver the terminal server apps would work for our environment, quest supports such things as app isolation etc. We were able to have a con call with a large quest implementation out of the states that re-assured us of the scalability of quest. Though xenapp has more functionality it isn’t worth the 350.00 to 400.00 a seat, and to date we haven’t required the extra functionality.

I have been fortunate enough to have access to all of the PCoIP bits since early on and as a vdi only user for the last 5 years I can say with confidence based our list of needs no one else is delivering as good of a product as teradici’s PCoIP(hardware, software or a combination of both)

My lab is complete with a

1.wyse p class(PCoIP) connected with multi monitors

2.samsung 19" with speakers(PCoIP chip embedded) multi monitor

3.samsung 19" without speakers (PCoIP chip embedded) multi monitor

4.wyse C class with multi monitor(software PCoIP)

They all work solid, deliver great video and usb support..use no more network bandwidth than any other vendor while delivering multimedia.

I have been testing the top 3 vdi vendors and the top 3 thin client vendors now for just about a year and I keep rolling back to the VMware \ Wyse solution for simplicity and stability.

With the introduction of the PCoIP monitors connecting to view the TCO has come down quite a bit on the desktop real-estate\financial side.

Feel free to email if you have any questions about our work to date..


can some please answser this question

if i want tio access view 4 with PCoIP remotely what do i need on the end device for connectivity?

Will PCoIP work with Vmware's SSL proxy/gateway?

Will it work with any of the SSL VPN solutions in the market or do i need an IPSEC VPN client becuase of the UDP traffic??