VMware announces Horizon Air Hybrid Mode, Horizon 7, and AppVolumes 3.0. Here's our full analysis:

VMware apparently did not take much a winter break, because in the past week we've seen several huge announcements. This article is going to cover the desktop-related news, but Jack has also written about Workspace One and the new direction it is taking.

VMware apparently did not take much a winter break, because in the past week we've seen several huge announcements. This article is going to cover the desktop-related news, but Jack has also written about Workspace One and the new direction it is taking. For this article, we're going to cover three big things: AppVolumes 3.0, Horizon Air Hybrid Mode (formerly Project Enzo), and Horizon 7. Let's dig in.

AppVolumes 3.0

AppVolumes 3.0 gets top billing here because the announcement came out last week. VMware, along with the rest of the application management vendors, has been on a mission to simplify application management over the past few years. With this release, VMware addressed many of AppVolumes' shortcomings and introduced several new features.

The most important addition is AppToggle, which which lets you deliver single applications from an AppStack with that contains multiple apps as opposed to delivering the entire AppStack. If you're not familiar, an AppStack is a group of applications that you deliver to users. The idea of grouping applications together for delivery is simple enough on paper, but if you only need to deliver a few of the applications in an AppStack to a user, you either had to deliver the entire AppStack anyway or create a new AppStack with just those applications. It's not hard to see that getting out of control very quickly. AppToggle alleviates the need to carve off apps into new AppStacks by allowing you to turn on/off applications. It doesn't just hide them, though (VMware thinks this approach can be exploited, which is a topic that I think would make for a GREAT BriForum session!), it actually prevents the application's files and registry objects from being merged with the base OS until the user is entitled to the application. The end result is fewer AppStacks and easier management.

AppCapture and AppIsolation are also important in this release. AppCapture gives you a way to capture applications individually into their own AppStack, then merge that new AppStack into an existing AppStack later. You can then use AppToggle to control which applications users have access to. When an application is merged into an AppStack with AppCapture (via command line), it looks at the application and the AppStack it wants to merge into and compares their files. If there is a conflict, AppIsolation kicks in to help resolve it. AppIsolation is basically an automated implementation of ThinApp, the big difference being that the isolation is only used when it's needed, not by default. In the past isolation was the rule, and integration was the exception. AppIsolation spins that around.

The remaining features focus on management across the board, even into other VMware products. AppScaling with Multizones enables you to pair repositories (basically file shares) of AppStacks with different instances of vCenter in different regions, which makes it easier to scale out your applications while providing a consolidated view across all those regions. Integrated App and User Management Platform is a middle API tier that they use to connect the backend to the new Unified Admin Console, which one of the things we got to see at VMworld. All the EUC products will eventually share the same management UI, which will tie back to a powerful backend via the API tier. If you recall Harry Labana speaking about a "cake baking" approach to management, this is the first iteration of that.

AppVolumes 3.0 will be available sometime in Q1 in one of three versions: Standard, which includes UEM and AppVolumes; Advanced, which adds in AppToggle, AppIsolation, and the scalability pieces; and Enterprise, which adds support for vRealize Ops.

Horizon Air with Hybrid Mode (formerly Project Enzo)

Announced today, Horizon Air with Hybrid Mode (which is a mouthful, and HAHM is not the best acronym, so let's just call it "Hybrid Mode" for now), is the long awaited cloud management plane from VMware. While it's not identical to Citrix Workspace Cloud, it does set out to make it easier to deploy the management components associated with virtual desktops and applications as well as the deployment of the desktops themselves. It leverages select vSAN-ready nodes that are placed on-premises that connect to the Horizon Air with Hybrid Mode service. Admins connect to the service to manage desktop pools, and users connect to it as their broker. Because it's a cloud service, it's always up-to-date.

Horizon Air with Hybrid Mode is not intended to work with traditional on-premises Horizon deployments. The service leverages Horizon Air, so if you use that on-premises, you could move your management to Hybrid Mode, but for the moment it seems that this is for net new deployments that use vSAN-ready nodes and HCI appliances. Future releases should see support for cloud services beyond Horizon Air.

Horizon Air with Hybrid Mode retails for $16 per named user, per month ($26 per concurrent user) if you already have licenses for vSphere, vCenter, and vSAN. Those prices assume that you already have VMware virtualization software, but if you don't you can add on a few bucks to the subscription to get them, too (bringing the per-user cost to $18 and $30, respectively). In either case, you'll need to buy the hardware (or Horizon DaaS) to host your desktops. Like AppVolumes, it will be available in March.

You'll note that even at the $18/per user, per month price point, Hybrid Mode is about half the price of Citrix Workspace Cloud. I wrote back in December that Citrix had left a lot of room for VMware to undercut them, and it appears they've done just that. I'm sure Citrix knew this was coming because at Summit they announced that CWC pricing would be reduced, though they didn't let on what the new pricing would be. I suppose they were waiting to see what VMware would do and what, exactly, would come with the package. Expect to see a new CWC price from Citrix sooner than later.

Horizon 7

Last, but not least, VMware announced Horizon 7, which is also due for release in March. In addition to a tie-in with AppToggle and AppVolumes 3.0, there are two big themes to come out of this announcement: Just in Time desktops and a new, home-grown protocol called Blast Extreme Experience.

Just in Time desktops is the name given to what you may recall as Project Fargo from VMworld 2014. In early August 2015, it was released as a PowerCLI Extension called Instant Clone, but this was more of an early-access experiment than a real product. It works by allowing admins to fork a running virtual machine (called a "Parent VM") rather than taking a full clone. The forked VMs are called Child VMs, and, because of the fact that they are leveraging the memory and disk of the parent VM, they can be created in just a few seconds (or even less in certain scenarios). Each Child VM has it's own delta from the Parent VM, and you can specify unique settings for each Child VM via a "PostCloneScript." Because a Child VM was forked from a running VM, there is no boot up. "Powering On" a Child VM makes it instantly accessible. More information on this can be found in the VMware PowerCLI Blog.

Today, we're finally catching a glimpse of the real deal. With this technology, you can now deploy and update images exceptionally fast without recomposing or re-balancing them. You simply update the master, and the next time a user starts up their machine the changes are there.

The other big news from the Horizon 7 announcement is that VMware has created a new protocol called Blast Extreme Experience. They do NOT call it the BEE protocol, which is good because I can picture all the cheesy marketing right now (Handing out Honey Nut Cheerios at VMworld, Shawn Bass dressing up in a bee costume, etc…). It's called Blast, and though it may seem like it's a brand new protocol, it's something VMware has been working on for several years, tracing its roots at least back to the Horizon View HTML Access protocol from 2013.

VMware developed Blast because they wanted something that could work well from the cloud as well as on-premises, and though PCoIP has done well in both scenarios (well enough, even, for Amazon to use it for AWS Workspaces), they also wanted something that would work well with mobile devices, be more battery friendly, and consume less bandwidth than PCoIP for video. To address those priorities, Blast uses H.264, which is already optimized for video and has hardware-based decoders on just about every piece of hardware you can think of, which gives them the battery life they were looking for as well as the ability to work on very inexpensive thin clients. According to VMware, users can work on battery power twice as long using Blast as they can with PCoIP on the same device.

Blast uses UDP by default, but falls back to TCP if needed. With UDP, they can handle 20% packet loss and still present a reasonable user experience. (That's subjective, of course, but it does show that VMware considers this a modern protocol for use on the LAN, WAN or on mobile networks.) There are no feature gaps between Blast and PCoIP. It supports MMR, USB device, Client Drive Redirection, and all the other features you'd expect, plus it's been developed with support for GPU Offload built in. Deploying Blast amounts to an extra check box when setting up a Desktop Pool and selecting the protocol to use by default. Older clients that don't support Blast will simply fall back to PCoIP or RDP.


I'm not sure what this means for the relationship between VMware and Teradici, though I'm inclined to think nothing will change in the near future. I'll be curious to see how the performance of both protocols compare in the real world before making any judgments. I do, however, think that it was only a matter of time until VMware made their protocol. Overall, though, I think it's a non-issue at this point.

The changes made to AppVolumes are critical given how the App Layering space is evolving. They've done a good job of expanding the platform without adding back in the same kind of complexity we've had for the last 15 years. Even the AppIsolation component only isolates files/registry entries if it has to, rather than by default, and the fact that it does it in a hands-off way is critical.

It's too early to tell whether or not Horizon Air with Hybrid Mode will be successful, but in general I think we're looking at the future of virtual desktop and application management, just as I do with Citrix Workspace Cloud. They are both taking different approaches, targeting different classes of users, but it's interesting to watch as it all evolves. If you're just getting started with VDI, though, it makes a lot of sense, especially if you consider that you're starting with Horizon 7.

Dig Deeper on Virtual desktop software and vendors

Join the conversation


Send me notifications when other members comment.

Please create a username to comment.

So a few things I'm thinking of here:

1. Does Horizon Air with Hybrid Mode require EVO:RAIL as was originally required with ENZO?  If so it's a non-starter for HPE.

2. VMware's blog says that BEE (heh, you shouldn't have said it in the first place) is optimized for Nvidia GRID.  I'm wondering how well it performs in comparison to PCoIP.  I'm especially interested to see how it does with 4K and multi display use cases.  I'm getting a lot of calls for GRID enabled VDI.

3. I can't wait to get my hands on AppVolumes3.  Even with the shortcomings of previous versions, AV has been great.  The integration with UEM should also help simplify things a bit.


@Rick - on #3 - love to see you and the community begin true side by side "bake-offs" but more to the point - "Use Cases" for each of the techs from @Vmware AppVolumes and @LiquidwareLabs FlexApp / ProfileUnity

I think that will really help parse out some of the technical, use-ability, scalability, security, compatibility questions that folks will have.

2016 - Year of Layering Folks ?????  


@Rick - You do not require EVO:RAIL as you did with Enzo. I was told that VMware actually delayed the release of Hybrid Mode so that they could relax that and allow you to deploy it to any vSAN-ready nodes.

(Also, I'm with you on Blast...can't wait to see it)


Interesting, so both VMware and Microsoft are moving their respective remoting protocols to use H.264 as Citrix has for some time.


@Gabe Just a brief double click on AppToggle and AppIsolation. AppToggle will still allow you to work in single app or multiple app mode if one desires and still have multiple AppStacks for the ultimate flexibility and simplicity of management. Also the reason we think AppToggle is more secure- If the apps aren’t installed in the operating system then by definition there is smaller attack surface vs. hiding.

AppIsolation. App Volumes since the beginning has always searched for dependencies within it's own AppStack first and ultimately looks in the WinSXS folder to honor native Windows functionality. The side effect of this is a soft isolation, which we call Volume Isolation where you can have things live side by in a volume or across volumes and not conflict. However at run time, when processes require true run time isolation , I.e when they are launched from things like svchost, we honor those standard Windows conflict and it that case provide Full Isolation with our ThinApp technology, which is integrated with AppCapture. I'd encourage people to understand this basic technical fundamental of how Windows works. If anybody has a valid technical retort to that in code, I personally invite you to come and join my Windows Kernel engineering team who are world class!

Also I'm personally really excited about the Just In Time desktop vision coming to fruition with Horizon 7. It has taken a lot of deep collaboration internally and App Volumes is the only solution certified to work with it. As you can see App Volumes is now a core technology, part of Horizon on-premises, hybrid, Air and also available standalone for Citrix customers.


Couple of takeaways from today's announcements....

1. Project Enzo (now HAHM) requires a specialized piece of hardware to be onsite to make this magic happen.  That means more capital (I'm sure it's not cheap) in order to connect to cloud control plane.  I find that to be really disturbing b/c if you have capacity in your onsite data center, why can't it just use whatever hosts / storage you pin it to, much like CWC?  Why do I need separate hardware just to run desktops?  Not feeling the love there.  

2. As for BEE, another steal from the Citrix playbook.  What would VMW do if they didn't have Citrix to steal from?  I do wonder how it compares to framehawk though.  

3.  I am excited about AppVolumes though.  



You make it sound as though specialization is a bad thing Nick.

To be clear the only 'specialized hardware' that Hybrid Mode requires is a hyperconverged platform from one of VMware's approved partners. I'm not sure that hyperconverged even qualifies as specialized, not any more. Perhaps 'optimized' is a better way to consider it. Right now it looks like the number of approved Hybrid Mode partners is quite a short list - Dell, EMC and Sphere 3D, but I fully expect the list to grow as HP, Lenovo, and Fujitsu start to catch up. And while I'm not sure what you would have to pay one one of Dell's boxes, I think you will find that running HAHM on our V3 workload optimized appliances compares favorably with delivering distributed desktops the old way.

Sure you can do it yourself. If you have over-built your virtual infrastructure and have the in-house engineering skills and time needed create your own platform optimized for desktop workloads, go for it. But if like many people you have a virtual infrastructure optimized for server workloads chances are you won't be able to squeeze in those full high, full length, double width, GPU boards that you really need to be thinking about today. So you are going to have to buy new hardware, and if you are doing that, better to buy HCI than anything else unless you are operating at very large scale.

I had doubts about Project Enzo when it was first announced; too much emphasis on the control plane, too little given to App Volumes and Just In Time Desktops. However, now it has been released I really think that HAHM is VDI done right. VDI has been struggling to make inroads into the enterprise because it was too complex and too expensive for most organizations to contemplate.

They just need to do something about the name...


It's good to see that the EVO:Rail requirement is gone. I wouldn't get to play with it at HP is that requirement had continued.  Is there any reason that a hyperconvered solution is needed locally (like our HC-250).  Why no a converged solution like our BladeSystem and 3Par? Is this just a marketing decision or a technical one?

@t.rex, I might be able to make that happen. But beyond that, my customers have been more interested in AppVolumes with UEM becase they are wrapped into Horizon's licensing. I'll hollar at you later..

@Gabe, thanks for that info. I've got a GRID K1 in our lab, I think I'm going to spring for a 4K display and give this a go in testing. Now the only real issue with GRID is Nvidia's new licensing model (you should an article on that).  

@Larry, y'all have appeared to have done a great job and I always look forward to the GA's getting to us.  Keep up the good work.. That said, have you guys figured on the use cases that will take the biggest advantage of JiT desktops?  IMO it is a "nice to have" feature vs a "must have" one. There are going to be additional costs to JiT-D (Copyright RickBoyett 2016 ;-)  )  and I'm not sure the gains justify those costs yet.  At the very least, I need good reasons to propose it to my customers.  

BTW, why doesn't Mirage ever line up their updates with the rest of your big product updates?  Just curious on that...

As for BEE stealing from Citrix, when did Citrix start offering an agentless HTML5 based client?  I always thought Blast was the first.


The "who came first" debate in software is meaningless IMO, unless you also have patents to lock the market. Otherwise the question should be "who better fits my needs?" Often it's not the product which came first.

That being said, I think the first HTML5 remoting offering was Guacamole, but originally that was only VNC. Citrix demoed HTML5 remote access with Google when Chromebooks were introduced, but then didn't follow up with an actual product for a long, long time. Ericom had the first enterprise HTML5 remote access solution (RDP based), followed closely by RemoteSpark. VMware demoed Blast almost a year later, but actually released it only another year afterwords, shortly followed (finally) by Citrix.

BTW it's funny that this protocol has the Blast moniker since, based on the description, I'm guessing it doesn't support the HTML5 client - maybe only with the TCP fallback. Anybody know? And does Blast support app publishing?


There was a VMware Horizon network map of the different server components and the ports each uses.  The agent for the Linux VM's uses only Blast.  I never got to play with that component.

I remember reading about Guacamole.  They collaborate with the FreeRDP project to get that feature.  Glad others began to notice their work.  I hated RDesktop, still based on RDP from NT 4.0 for Linux . They're still hard at work adding more features to it.  HP now uses FreeRDP in their Linux thin clients now.

Citrix has come a long way with their HTML5 client.  It can play audio.  The ICA protocol is pretty responsive.  local devices mount in an ICA session.  Can access local drives and Google Drive.  Printing I think you need AirPrint/ePrint to work however.  I was hoping for a WinRT variant with Office 365 integration.