There's been lots of news and good discussion last week on layers and user virtualization, both topics that I've previously discussed here and here. I've been reading the various posts and blogs and it got me thinking about what this means for RDS and XenApp in the context of desktop virtualization?
The RingCube acquisition is the first nail in the coffin for XenApp published desktops. It's clear to me as I look at the world through my lens that "desktop in the data center" is the lead message for everything Citrix does. (Ignoring the online and networking business). But when it comes to desktop, the current message is all about desktop as opposed to "apps" and "access" as it's been in the past.
XenClient is still a toy and will not have broad application any time soon. This desktop-led message has been a great marketing exercise for both Citrix and VMware, who are both selling a promise of a better desktop. However, who's really deployed anything? As we know, in this industry (outside of some who have gone 1-1 VDI and brute forced deployments) most people step back to XenApp based desktops and a mixed physical environment. Nobody can claim that mainstream adoption is happening.
How has Citrix reacted to this? Their first step was to sell XenApp trade-up licenses which has worked very well. But listening to the analyst calls every quarter it became clear that the questions started to get harder about what represents XenDesktop sales vs. cannibalization of the core XenApp business etc.
In reaction to this, Citrix provided some more fluffy information on VDI breakdown vs. broader Flexcast (the mix of all the Citrix desktop offerings) sales under the desktop umbrella. This gives the impression of growth, growth, growth of the desktop business. Cannibalization of the standalone XenApp business can be explained away due to growth in the desktop business, which includes XenApp, and this makes sense.
This has been a great story to date. Customers are being given great deals on trading up XenApp licenses and opportunities created for resellers to go back into customers who may be sitting on expired licenses. XenApp has been the killer feature in the broader XenDesktop suite that is letting Citrix beat VMware with their "VDI only" message. This strategy has certainly sold lots of licenses, but money is also made by renewing maintenance and to do that customers must start to implement.
You can only trade up so much before people see through the smoke and mirrors and realize they have lots of licenses that are not deployed and therefore will not renew maintenance contracts. This is also why I suspect Citrix is starting to talk a lot about desktop transformation at Synergy and seems to be continuing the thread at Synergy Barcelona, as evidenced by the online agenda. As people start to implement they will also start to realize all the limitations of XenApp as less flexible versus a 1-1 PC.
So, a more flexible desktop must be created. Can you do that with XenApp? Sure, but not with layers today or anytime soon. This means the only option to make a better XenApp is "user virtualization".
But instead Citrix seems to have chosen to focus on VDI and to acquire a simple layer technology in RingCube. By "simple," I explicitly mean less capable than Unidesk. I believe this represents a fundamental shift in strategy. Flexcast is too complex in reality for many. I find it ironic that Citrix CEO Mark Templeton said in his Synergy keynote that "complexity is optional" and in reality Flexcast is just a bag of unconnected technologies that exemplify complexity.
That may be changing, however. The acquisition of Kaviza for low end VDI-only solutions will certainly appeal to many, and could even become a better 1-1 model for large enterprise in the future. Citrix has XenClient, Kaviza, and enterprise XenDesktop and XenApp. RingCube, from what I understand, could work across all of those models, except XenApp due to the multi user nature of RDS. Having a simple layer that goes across these different architectures could help unlock a lot more adoption and simplify the consumption of these different models. However there will be much resistance to get to this supposedly simpler world with XenApp published desktops still an easy choice for the vast majority.
So how can Citrix fix this?
You can start with a layer to abstract applications and personalization. Both Harry Labana and Ron Oglesby are right when they talk about departmental installed apps and personalization in a layer respectively. That layering, if it doesn’t work on XenApp, will start to get traction and resources redirected to the VDI only model. Assuming it works and keeps improving, why bother back-porting it to XenApp to slow down progress? Why not just go faster and make it better so it's very close to XenApp in terms of price/performance? That doesn’t mean kill XenApp--just refocus XenApp to become an application-only story or even better perhaps evolve XenApp to deliver new application types as a future cloud app delivery platform as Microsoft continues to make RDS better. (I guess this is why Citrix invested in Gizmox and also created their startup accelerator fund to explore new opportunities.)
After thinking about this for while, I started to factor in the latest organizational changes at Citrix that I've heard from various people which suggest that big changes are taking place. I saw an organizational chart of the new California-centered Citrix and it is worrying. The fired previous GM of XenDesktop has been replaced by some HP dude with a storage background, and the leadership for the whole desktop and cloud space is all in Califorina under, amazingly, the previous guy who ran the desktop business (who let's just say in IMO has done little to move things for those of us trying to deal with the realities of desktop implementation.)
The only good news in the new organization is that there will be a new CTO role reporting directly to Templeton. That will hopefully make things better over time, however, I see no impact for some time. Couple that with my views on XenDeskop product management, and that nobody who understands XenApp or the desktop in charge, and it's worrying. At the very least the organization is now in place that does not care about XenApp and all the pieces are in place to go full force ahead with enabling VDI.
But will this strategy work?
I believe it will certainly reduce the need for XenApp published desktops. Kaviza or XenDesktop enterprise with a single layer would be a good fit for a simple organization, so it's not all bad news if we get some focused effort. However the RingCube layering technology has holes that Ron Obelgsy certainly is keen to point out. But, using Ron's guidance we would end up with multiple layers, even per app! I'm not convinced it's going to reduce OpEx complexity, solve application compatibility, be simple to deal with storage in the datacenter, or scale. I also think it will confuse many people who use System Center and/or App-V, and it expects customers to trust that their vendors will support their applications in a layer.
This is way too much for the average customer to chew off and represents too much disruption to adopt especially when the Windows 7 migration deadline in 2014 is fast approaching. Even for the complex enterprise its a huge bet to make. I only need a few apps not to work in layers, or to not tolerate layer conflicts to totally break my model. It's a bit like application virtualization, which for most will be in addition to electronic systems distribution (ESD)--not a replacement--since App-V does not work for all apps.
Layers will make this problem even worse if you create too many of them. So with respect to layers, I do think Citrix did the smarter thing by taking the simpler, less risky approach. The other thing I don't like about the layers solutions is that they don't solve user management use cases. Sure, they can offer some personalization and profiles, but that's not the full story that is sometimes implied and more often confused based on my various readings. Sure you can customize a desktop with a layer and put things in there, but it has nothing to do with management of the user layer. They are as, Brian articulated really well last week, two very different spaces. Both are needed to truly build the layer cake. Its a lot more than profiles, stupid! :-)
So if too many layers is confusing, VDI only (i.e. no XenApp) will not be ready for the masses before the Windows 7 migration window runs out. Why is Citrix betting on a technology that does not strengthen their Flexcast message? I've been scratching my head over this one. I have two theories. The first one is that RingCube was a fire sale, although that’s impossible for me to be certain of since Citrix did not disclose the price they paid. I have heard through the rumor mill that it was around $50M-$60M, so that’s pretty cheap in the grand scheme of things.
Let's face it, what was RingCube going to do? They have been de-emphasizing their virtual workspace client hypervisor alternative for a while. I doubt Citrix will continue with that to avoid confusion of the XenClient message, but perhaps this is exactly what they will offer in addition to an app/personalization layer to compete against Type 2 solutions while claiming that XenClient (insert marketing term to describe RingCube based desktop) now runs on any desktop that can run a Type 2. This would also ease the transition to Type 1 since the management model would be the same. I would love to see Citrix do this. My second theory is as I suggested above. A shift in strategy with a new organization focused on VDI only that will evolve XenApp back to apps over the coming years. In other words Citrix wants you to deploy VDI only for your desktops in the datacenter.
Where does that leave us? A solution that is not really going to help us much in time to complete our Windows 7 migration. It's VDI only, does not deal with multiple layers, and lacks a management model to deal with all the complexity. So what is the value? The more I think about it, I think departmental installed apps is exactly what it is targeted for. A simple solution designed to help more people migrate to the VDI only model and a way to deal with groups of apps. This will leave those who bought into the Flexcast message and on XenApp based desktops hanging.
The message, I predict, will be very simple. If you are an SMB player and don't understand RDS/XenApp and just want VDI for the sake of it (or just want to implement VDI without having to change your management model), use Kaviza. If you want to start down the path of a simpler management model, use the RingCube layer to start to group departmental apps so you can manage the base image separately and have the option of some personalization within your existing distributed management paradigm. If you really want to implement Flexcast with XenApp and offer user based management then trust us long term, use UPM for now (of course they won't explain that it's profiles only) or go speak to our eco system partners. This I guess also achieves the goal of keeping their eco system alive, which is less risky than trying to say Citrix can do it all.
What do the remaining layering vendors do?
This presents an interesting dilemma for the remaining layers vendors. Unidesk in particular are in trouble, IMO. They represent changing everything to their way...storage, management, apps etc. They don't really play nice with anybody. A brave strategy. I am not trying to attack Unidesk or Ron, it's just my opinion. They are good, smart people, and I like what they are attempting to do. However, VMware may conclude that if RingCube got bought cheap, Unidesk has nowhere to go, so they’d be a good technology to pick up with smart people and a better counter-feature tick box in XenDesktop comparisons. They have the option to strip Unidesk down to make it simpler like RingCube if they need to compete on simplicity.
There is also a play here to try to scale Unidesk. RingCube won't be able to scale as well as Unidesk because they don't have all the storage smarts that Unidesk does (although if Citrix does tricks in XenServer to counter this it would be no surprise). VMware could also try to solve all the application compatibility problems, integrate ThinApp, and try to go head to head with Systems Center. Would VMware really want to do this? I'm not sure they would because this would mean keeping customers in a Windows world (and clearly that is not what they are trying to do).
Another option is go pick up somebody like Moka 5, who would be really cheap from what I hear. This would not be for the Type 2 client ypervisor solution, although they could use that to disrupt XenClient and complement VMware Workstation with a better management model. It would be more for the layering technology that Moka 5 has. It would also be a blow to Quest, who I guess have been picking away at the View market with their improving solution and Jon Rolls leading their partnership with Moka 5. Wanova may be another one, but they’re a pure management play that has nothing to do with the hypervisor. I believe Wanova suffers from many of the same problems as Unidesk in that it’s their way or the highway (AKA too much to bite off for the average customer).
After thinking about this for sometime, I came to the conclusion that a better layer solution is not going to mean much for VMware. I then turned my attention to considering what would happen If instead VMware decided to take out somebody like an AppSense. Plenty of large XenDesktop customers use AppSense for VDI. VMware is never going to get into an account trying to win the protocol battle, so why not change the desktop conversation more to what the core of their business sells and is trying to achieve.
VMware is becoming a management and security company as far as I am concerned. Imagine if VMware could claim they are the best management layer for Citrix Flexcast :-) That could really hurt the price value of XenDesktop and legitimize VMware in the desktop stack in the accounts that are leading the move to VDI.
What's even better is that AppSense already has a layers technology that they haven't brought to market. That could become a competitor to RingCube, assuming it is any good. VMware could own the user management stack which is used today in many large accounts and move into layers, instead of the other way round, which would give them something to differentiate them from Citrix. Management of the user could also lead to models that move beyond Windows use cases, and with policies could really complement project Horizon.
All of this could be applied to both physical and virtual environments and could help capture mind share in the enterprise and help the move to non-Windows-based apps. VMware marketing talks about modernizing the desktop, and this could be a way to get there starting with today's desktop. As Brian Madden says, moving from the Big D desktop to the small d desktop.
Thinking about it some more, it would be a disaster for Citrix, who would then have no choice but to take out RES who's technology is not used by anybody at scale that I know. I could go on to express what I think of RES, but I will reluctantly restrain and not go off on a tangent rant :-( .
Objectively, since RES has next to zero presence in the large accounts due to inferior capabilities, the strategy of reacting and taking out RES would not really help Citrix much and they would be left holding the bag with a weak offering in layers and user virtualization. Personally I would like this because I could then use that as leverage to negogiate a cheaper renewal with Citrix, Microsoft and also avoid getting too locked into the Citrix XenDesktop stack. I like how today I VMware for the hypervisor, Citrix for the protocol, and different management tools. Management is a gap for me moving forward, and I am unlikely to write advanced user based management tools myself. I’d rather integrate parts of them instead of deal with the entire thing. Actually, I'd like to see the layers and user virtualization vendors enable APIs that let me write to their technology so I capture what's important to me, as opposed to having to buy into their whole stack (which is a much bigger leap of faith and barrier to adoption since I also worry about becoming locked into them).
In Brian’s post, Doug Dooley from Cisco commented:
"I'm also cheering for VMware to make some similarly aggressive and innovative moves to take us beyond this boring, corporate Wintel homogenous PC-world most of us live in. Let us get to a world where the "P" comes back to the "PC" at work and the "any device" vision of apps, data, and a rich collaborative experience is real for the common person."
I find it interesting that Doug is giving hell to VMware so he can fight a Windows only world, all while Cisco and Citrix are becoming buddies... VMware has a chance to screw up this partnership, a partnership that does not help them. So I think Doug's challenge should be taken seriously by VMware.
If VMware does nothing and still pretends that a delayed RTO profile-only solution (Persona) is even worth a discussion at VMworld then shame on them. Every customer should get off View, give up on VDI or migrate to a competitor if that happens. If VMware chooses a more robust multiple layers solution, is there a future that supports their strategy? Maybe, and its still better than doing nothing. If VMware leap frogs Citrix with user virtualization, it gets access to enterprise accounts in new ways and creates price leverage for customers today while disrupting the Citrix/Cisco relationship (and I am sure many others).
Despite this, I don't think VMware has the will or mindset to do much, so I think they will play it safe. Everybody I talk to says they are arrogant, not just me and my experience. They think they know best when in reality they don't understand the desktop. Fighting the protocol war just shows how weak the end user computing strategy is and they clearly need a new strategy. The question must be asked, has VMware given up on their desktop strategy and are they trying to make a graceful exit?
There are opportunities for them to compete head to head with Citrix in new ways today (which they are trying to do with Horizon) or even to Judo flip Citrix's own message because Citrix is focused on VDI-only moving forward, led by weak leadership who are talking about PC 3 when they can't even get PC 1 right. As customers we need VMware to compete and keep Citrix honest. It would be a huge loss not having a strong competitor to Citrix. We will continue with poor execution, lots of vision, and snail pace innovation from Citrix. So come on VMware, act and do something quickly before it is too late, or tell us now to get out for the customers sake.
Update August 26
So two weeks later do I feel much different? Not much. VMware has let leak View 5 documents and it’s clear that they are still only at the basic profile integration stage. Ok XenClient 2 is out with the annoucement of XenDesktop 5.5 and it’s good to see it progressing from toy to the next stage with better management something Citrix could have done 2-3 years ago if they had the balls to buy Moka 5 instead of snail paced management iterations as we now have to yawn about for several more years. The HDX improvements are welcome and further underlines why I don’t think VMWare can win that argument for a long long time.
What has transpired though is a reaffirmation of my believe that the big eco system vendors are taking the simple is good enough for layers as posted by Ken Bell of Citrix here. It also confirms to me that layers are focused on apps at least to start with. I also saw the Liquidware labs announcement today regarding FlexApp. I also saw the Liquidware Labs FlexApp announcement. As I read this article by Bridget Botelho which is also linked to on the Liquidware Labs home page its apparent that this is her opinion not a formal VMware endorsement that says this is their answer to RingCube. However the approach is once again a simple layer approach as opposed to what Unidesk offer. Watching some of the Tweets flying about, it's clear Ron from Unidesk is keen to point out that layers are not that easy and a lot is involved. Outside of the technical arguments, I believe this is also a battle of minds that is emerging. A simple layer for the right use case or a comprehensive new way to manage the desktop with layers for VDI? The Appsense take on layers via Harry Labana is also interesting and for me two key points stick out-ignoring the RES barb which I enjoyed finally seeing somebody other than me finally standing up to them and exposing them for what they really are :-). Appsense is clearly indicating their intention to have a layers component to their feature set and I think the question about will the apps work is the right one to consider.
This is a key question to consider if you want go down the layers path. Microsoft is silent on the matter. Perhaps because app layers would negate the need for App-V in many cases and reduce it to an isolation or on demand app style use case only. That all depends who you want to bet with. Microsoft who own the operating system or some layers technology that you hope will get enough traction that the vendor eco system will support. I think its pretty clear by now what my position is on layers, a niche technology for some use cases that will take years if ever to mature that is VDI only so only goes so far. Unless I hear a direct endorsement from Microsoft I remain skeptical about its chances as they always want to control the base platform execution environment. Perhaps that's where Citrix is best positioned to influences Microsoft, time will tell. Going back to Liquidware Labs, I am not sure if they can succeed either. They are hiring more Sales people and they claim to do everything, assessment, profiles, layering really? How good can they really be and is this just more marketing from a small vendor that really has no future? Even in the assessment space it's interesting to observer that Lakeside has now formed a partnership with Appsense. So will be interesting to see how Liquidware Labs will try to compete. If all they do is put price pressure on Appsense etc by offering a low end alternative then I am fine with that. VMware is next week, so let's see firstly if I can get published this time and secondly if VMware will make any statement about how to respond to RingCube.
I maintain responding with just another feature is a weak position, what they need to do is take it to the next level and I don't believe they can win with complex layers even if layers is part of their strategy. Looking forward to hearing the news as I take some time off.