This post started as a comment on Brian's article titled "MokaFive plans to release 'Layering' as its own product. Nice!" but I think this topic is important enough to warrant a blog post with it's own commentary.
First, I want to say that it is refreshing to see from Brian's post and the ensuing comments that there is a good understanding that Layering and User Virtualization are different focus areas with some overlap. Both are required to enable the layer cake. It's amazing to me based on my interactions and friends' interactions how poorly both VMware and Citrix understand the problem within their product teams (minus a few exceptions). For the most part the product teams that I have interacted with have no clue what's involved in real-world desktop implementations and are led by clueless executives who just pander to customers with generic statements and do nothing.
Case-in-point: we've been talking about the layer cake, stateless desktop, etc. for years, and the leaders in the desktop virtualization have done nothing to enable it except present PowerPoints and marketing statements. I think it's fair to generalize and say that, in effect, neither provides anything in the layering area. Sure they will argue back and say their respective app virtualization solutions are a part of the layer cake. True, but it's really a coward vendor answer avoiding the meat of the problem that needs to be solved. Citrix will then probably go on to argue that they also have that junk lowest common denominator solution that they licensed (or whatever) from Sepago.
VMware is even worse and does nothing for users thanks to RTO. Interesting question here. Is it because RTO sucks or is it because VMware would not agree to a sweet heart deal with Symantec as a condition of the acquisition of RTO? To refresh your memories, Symantec used to distribute RTO as their profile solution. I've heard different theories here, curious if anybody knows the truth. The party line I've heard is that RTO sucks and they found out when they tried it on Windows 7.
I personally smell a rat here. I've never been a fan of RTO, but surely it can't have been that bad? Regardless, it's a huge gap for VMware. Also, I don't see anything in their latest earnings report that show any signs that their desktop revenues are relevant. That means they will not invest in the desktop, which is bad for the industry since it will make Citrix even stronger and slower to innovate than they already are.
As a result, the world will continue to implement desktop virtualization in one of two ways. The first is 1:1 desktops for the VDI model which is expensive and complex. This will however get better over time with greater core density, storage technology, etc... It won't be mainstream for years, though (if not ever). The other alternative is RDS based solutions like XenApp/Quest. These have their own set of limitations as we all know and love :-) If anybody is actually implementing pooled VDI desktops successfully as opposed to RDS I'd love to hear why you do and are willing to spend so much extra money for it.
So where does that leave us? Desktop Virtualization is a niche and will remain that way. The only way to take desktop virtualization mainstream is to enable the layer cake. Microsoft will tell us to make the desktop cheaper with Application virtualization and Systems center, but that is only part of the problem. Application virtualization does not work 100%, is not a standard, and therefore only gets us part of the way there. Systems Center is too big and complex for many people and over priced as a result. That's why I guess MS will use InTune to address simpler use cases.
I'm confident they will screw up both of them because they will approach both from a distributed computing mindset. The whole layering/user virt thing requires a different mindset involving central management and single images. App virtualization is the closet thing MS has, but even that has no pull within MS or developers. What about ecosystem support? How many vendors support or ship their application as an App-V package? Very few. That may change in the future, but the point I am trying to make is that all this has a very long lead time and there is a lot of uncertainly.
The question on my mind is, will layering ever work? We know from history that App-V will never become a platform since it's been around for 10 years and still 100% of apps don't work. The strategy from MS is to make money with it from MDOP and SA in the desktop world as opposed to making it an operating system feature. It may become a migration tool for legacy to Azure, too. With that in mind, what chance does layering have, and does that fit into the MS view of the world? Is System Center ever going to offer layers (since that is Microsoft's only management tool)? I doubt it. Certainly we will see very little of it in Windows 8 based on the leaks I have seen, and who really cares about Windows 9 this decade? By then will layers matter if the world is starting to move beyond Windows (I'll call it "classic" meaning today's model) to cloud something?
It feels to me that the problems will be different. Access to applications, data, settings, etc... (layers as we think about them today) will be relegated to classic Windows only without really addressing the future. So my theory is that layers, while needed, will take years to get right will need to be provided as a FREE feature of View/XenDesktop/vWorkspace etc. Even if that happens, the management requirements will be new and lots of change will happen. Desktop virtualization vendors will have to ask themselves if they now desktop management vendors. I can see VMware saying "Sure, that's part of vSphere. Look how valuable we are." I can see Citrix/Quest saying that the godfather in Redmond will never allow that, and asking themselves "how can we remain Microsoft's favorite pet by SLOWLY adding value on top of System Center only (or shall we go find a new master)?"
The bottom line is that I have very low confidence that layers will matter. By the time they do, the world will need to solve a new set of more compelling problems. That doesn't mean that people like Unidesk, Moka 5, Wanova, and Ringcube won't keep trying anyway. They should still do just fine, too, as there is real value in solving problems for market segments. As for layers changing the way the world does systems management, though, I'm not holding my breath. Our only hope is that the traditional systems management vendors decided to move into the space and give MS a kick in the balls. Until then, my opinion is that layering is just good blogging material (at least for the most part).
One company that I think has the ability to do the kicking is MokaFive. Those who follow my posts know I have been a huge fan of their management and security policy on client hypervisors for a long time. While I still think they have a good product for that use case, I also think NxTop is catching them fast. With regards to layers, I worry that they are moving away from their core focus. They are small company with limited resources. While I understand their strategy of trying to increase revenue with a broader appeal I am skeptical of their chances of success without more support like they received from Quest. I'm not trying to be negative--just pragmatic. I hope they don't lose focus. I would really like to see Moka 5 become a management layer for Type 1. They have solved a lot of the complex enterprise use cases which is why I tweeted that I think here is an opportunity for Quest to add value to XenClient.
I also see the layers camp dividing. Looking at the upcoming Citrix Synergy solutions expo list, it's interesting to note that both Moka 5 and Unidesk are missing from the list. Clearly Moka 5 is now in the Quest camp, and from all the various marketing I have seen it seems that Unidesk is mostly trying to play in the VMware camp. Ringcube has been included in the Simplify Desktop Transformation with Citrix Ready Partners section in addition to having their own booth. Wanova also has a booth. So will one of these players end up in the Citrix camp? Another unrelated observation is the lack of Quest at the expo, while VMware and Virtual Bridges are there. I wonder what reason those not attending will acknowledge :-) In my opinion, the whole ecosystem is up in the air. Alliances appear to be forming, and with all the uncertainty I outlined above I am not about to mortgage my management strategy on anybody.
So what is one supposed to do? I've been thinking about this lately. I firmly believe we have to move beyond a distributed Windows world to a single image management model at all layers, OS, Apps and User. The OS can be dealt with via desktop virtualization in many cases with a single image. Layering would be useful to patch just the golden image and much less risky in terms of breaking stuff. I can deal with that use case for layering. Apps can, for the most part, be virtualized. The ones that I can't I'd rather bake into the golden image and avoid trying to manage all the layer complexity, not to mention potentially avoiding app compatibility due to layer merging issue.
Even if I got this to work for many apps, I'd just need one app to break the model, have to move it into the base image, and once again have to ask myself whether or not it's all worth it. I'll also repeat my concern about getting support for my apps especially from third parties. I can barely get these folks to talk to about App-V, have spent years getting them to support XenApp, and I don't really want to start a conversation about supporting layers unless there is a compelling value.
That leads me to the user. It seems to me that the more time I spend trying to manage my user environment, the better it is for all my desktops and the better my chance to move to desktop virtualization where and when it makes sense. This also helps me reduce the number of images I need due to personalization needs. The more I understand about my users environment, the greater my chances of moving to future models of Windows. As the world becomes less Windows classic centric, new areas of user virtualization opportunity will open up for my solutions to evolve into. Where will the layers people go? Will they be relevant?
Lot's to think about, not sure I have reached a firm conclusion. My current view is that we have to have better/new desktop management beyond the current status quo. I see layers as hugely speculative with no major vendor commitment and a limited useful life of the solution. Therefore I plan to spend most of my time thinking about user virtualization and how to best combine it with some aspect of layering that helps manage just the base operating image. I'll also have to just deal with App-V more as Microsoft is pushing in that direction, and get my operational costs down to justify the expense.
I will also stay away from Systems Center for as long as possible, and I'll put more pressure on the desktop vendors to offer some layers golden image update capability in their base product to eliminate Systems Center. Until then, I'll stick to very basic App-V management or instead use ThinApp and run as many apps as I can from a network share. I think if I can do those things I have the highest probability of implementation success. The layer cake is needed, but the question is "where to start?" I think virtualize apps, user virtualization, and a single desktop image is the least risky strategy. If that's the case, will layers as touted by Moka 5, Unidesk, etc... matter beyond a niche?