BriForum 2014 London kicks off today! (If you're not here at the show, follow the action on Twitter and Flickr.) One of the cool things we did this year was to ask attendees when they signed up what question they'd like to ask me. So here's the list of all those questions, plus my answers. I actually wrote this straight through without editing or stopping. So these are my real unfiltered thoughts. (Though, to be honest, that's how I write most of my blog posts. :)
So here are the questions asked, along with my answers, in the original order from when attendees signed up:
Application virtualisation is becoming increasingly popular, but getting an application virtualised can be painful and takes considerable time and effort. How do you see this process evolving?
The problem (challenge?) with today's application virtualization products is that they were designed assuming all apps were bad and they isolated everything. (So, their main purpose was for isolation.) In more recent years we started using app virtualization as a deployment tool, especially for delivering apps on-demand on a user-by-user basis into shared, non-persistent disk images. This is a problem because while we love traditional app virtualization for it's on-demand abilities, we have to sequence every app to get them into the app virtualization environment, and that (as you point out), can be painful and can take a lot of time and effort.
So recently I'm liking the idea of breaking the app isolation apart from the app delivery. This is where more modern products like FSLogix and CloudVolumes come into play. You can use them to package apps without the pain of sequencing, and then you can deliver them with whatever tool you want. (Well, that's speaking broadly. FSLogix creates the isolated (but not virtualized) app packages, and CloudVolumes creates VHDs for each app. But the concept of separating the packaging from the delivery is the same in both of them.)
By the way this is not to suggest that app virtualization products like App-V are bad. In fact you can (and should) still use App-V along side of FSLogix or CloudVolumes to virtualize problem apps. The key is that with these modern approaches, you're only actually sequencing and virtualizing the problem apps. For the other 90% of you apps which don't have conflicts, you can deliver them on demand without the pain of sequencing.
Are Citrix a dead man walking? (Also asked by a second attendee: Are Citrix in trouble with the furor surrounding the XenDesktop/XenApp licensing model and the release of View 6 from VMWare?)
Man it sure seems like it at times. They have pissed off a lot of customers with the whole "XenApp is dead, no wait, it's back" thing, they have a CEO who's retiring without a named replacement, and their 17-year monopoly on single app publishing from RDSH is drawing to an end.
That said, I love Citrix's hybrid cloud vision of Citrix Workspace Services and their ability to extend Microsoft's Azure RemoteApp. Citrix did almost $3B in sales last year, with a vast majority of them on XenApp and XenDesktop. I don't see them going anywhere anytime soon. And the new competition is just going to make sure they stay focused and get all the features we need into XenApp 8+ and CWS.
Because of the complexity of creating a solution that supports all use cases within the typical company, do you believe restricting yourself to the smallest number of vendors offers greater advantages than choosing the best of breed from multiple ones?
This is a really great question. With the consolidation of vendors we're seeing today, it's almost impossible to not choose vendors that don't have overlap at some level. That said, a lot of the mega vendors have broad coverage because they bought smaller vendors, so you don't really have less complexity because not everything is integrated even if you just choose a few vendors.
I don't mind overlap at the vendor level, but I don't love overlap at the product level. Take VDI for example. I feel like you should pick VMware Horizon or Citrix XenDesktop/XenApp. If you try to support both then you're dealing with different portals, different clients, different features, etc. That would suck. Just pick one.
Same for virtualization platforms. I don't care whether you run XenDesktop/XenApp on vSphere, but I'd rather see one hypervisor for everything, one VDI product for everything, etc.
Did VMware miss the boat?
Depends on which boat. :)
Given the aggressive positioning of Citrix and VMWare in the VDI market, do you think there are any risks involved in taking on Dell vWorkspace at this time?
It would be tough for me to recommend someone taking on vWorkspace for a new build now. I don't expect that it's going away anytime soon, so if you're committed to vWorkspace then I'm fine staying the course. But for a new project, man, I don't think I could go down that path.
GPU Virtualization is one of the hottest topics right now. While most people seem to focus on GPU virtualization for high-end 3D graphics, it also seems a good way to bring graphics up to a decent performance level for "standard" users. Would you agree, and if yes, any recommendations on how to get there (choice of products, hardware, positioning in a general mix of VDI/RDS . . .)?
Citrix XenDesktop on XenServer is a bit more advanced than Horizon View on vSphere now, although VMware has announced their direction and they'll get caught up at some point.
The thing I remind people is that even though NVIDIA is leading the marketing charge for putting GPUs in desktop virtualization environments, remember that AMD (via ATI) has solutions here too, so don't forget to check them out.
Also if you're here at BriForum, be sure to attend Thomas Popplegaard and Magnar Johnsen's session How to be Successful with GPU Virtualization. Then ask them. :)
In light of VMware's recent announcements around Horizon what would you advise Citrix to do next?
I assume this question was posed before Citrix's Workspace Services announcement at Synergy two weeks ago. Luckily they didn't ask me what I thought, because I wrote that they should get into the DaaS hosting space.
But the real answer is that they should do exactly what they announced with Citrix Workspace Services. A few weeks ago I wrote that it was the future of XenApp and XenDesktop, so if you haven't read that article, take a look at it now.
Interested to have a look to the BYOD strategy of a leading European university?
Yes! Hit me up at the show and let's talk.
RemoteApp provisioning has, in my opinion, one major flaw with regards to support—the application almost always has to be installed on a server OS. User applications are almost always never designed to run on server OSes and are often unsupported by the vendor, so running into major issues with a major application can suddenly make your entire thin client estate seem like a very expensive problem. Do you think leveraging desktop OSes as part of the RemoteApp infrastructure would be a good future development in order to help maintain application stability/support and therefore make Thin Client/Remote Access less of a risk?
This is nothing new, since we've been dealing with this with Terminal Server and RDSH for 17 years.
For me it's not as much of an issue of app compatibility, but rather the fact that Windows Server OSes have different patch cycles than client OSes, so it can be tough to keep everything "the same" in your environment.
For years I advocated that if you want Windows in a datacenter, you should use RDSH for non-persistent images and VDI for persistent. My feelings about that were mainly around scalability (since you could fit more users on RDSH) and manageability (fewer images), but really that's changing now. In today's world it's easy to create a single master image based on a client OS and to then spin up as many as you need on-demand for non-persistent VDI. And products like FSLogix and CloudVolumes make it easy to install all your apps into a single image which you use as a baseline for everyone.
So I'll say, yeah, in 2014, you have options for using non-persistent VDI for everyone which will essentially deliver the same manageability as RDSH with decent performance, so if you have a risk of an app not being supported on a server OS, then this is definitely worth considering.
Say we go 15-20 years into the future. Do you think most users (including at home) will not have powerful machines with local software anymore and just run everything remote on the cloud?
TS, VTS or VDI? :)
It depends. :)
What do you think of the latest Citrix XenClient solution, do you think it's finally enterprise ready? Also, are there any other vendors that pose as real competition with XenClient?
The key is that with XenClient is that it's a bare-metal destructive install. So this is really only an option for company-owned laptops. I'm not familiar enough with it to know if the current XenClient is any more enterprise-ready than previous versions. Certainly Virtual Computer (the company Citrix bought to create XenClient) has been around for six years now, and people who use it seem to love it.
What's more interesting to me is the Type 2 client hypervisor solutions which are great for BYO use cases. I love what Citrix is doing in this space with DesktopPlayer for the Mac (and they announced DesktopPlayer for Windows which is supposed to come out later this year).
As for competition, you've got VMware Horizon Mirage which is not a hypervisor but gives you some of the same management capabilities. As for Type 2 client-based solutions, you've also got Fusion Pro, Parallels Desktop Enterprise, and Moka5.
What is next to virtualize since there's not much left now with user virtualization/environment management?
The virtual networking (software-defined networking, or SDN) seems to be a pretty big deal. I love the idea that link networking policies to VMs, so as you move VMs around the networking policies go with them.
It's also cool that SDN can break the hardware link to expensive networking gear. So just like today's storage virtualization vendors mean you don't have to buy expensive SAN hardware, SDN will mean the same for networking.
What is the biggest mistake you should not make?
I guess this question is coming in the context of desktop virtualization? :)
For me there are two things:
First is the whole "trying to do too much" thing. I believe that the single #1 reason VDI projects tend to fail is not because of the "VDI-ness" of them, but rather because customers try to move to VDI *and* they also try to move from the crazy current world of every user having their own images to a locked-down non-persistent shared image world for VDI. That's really hard to do, and scores of VDI projects have failed because of it.
The other biggest mistake is people thinking that they need a new technology to solve a problem which they can solve in other ways. There are million variations of this in IT today, but I'll use DaaS as an example. Some people choose DaaS because they want to "outsource the management" of their desktop estate. Some DaaS providers do nothing more than giving you a blank desktop for maybe $50 a month, while others might charge $200 a month for "full management." So people think, "Great, I want full management, so I will pay $200 a month for fully managed DaaS." The problem is that then you have DaaS, which is VDI, and maybe VDI doesn't make sense for you.
If you just want someone to manage your desktops, you might be able to find a service provider who offers full management of your traditional on-premises desktops for just $150 a month. So then you solve your actual problem (wanting someone else to manage your desktops) without introducing new problems stemming from the adoption of a new technology (DaaS, in this case).
What is the Future of Workplace? (Another person asked: How will EUC workspace look in 2020?)
Since these questions came from the BriForum 2014 London registration page, check out the session from Ruben Spruijt and Jeroen van de Kamp called CTO's Perspective on Workspace 2020.
What is your opinion that more and more vendors come with there own PowerShell cmdlets, making this the new universal way of system management.
Yes yes yes!!!
The best part about this to me is there are so many ways to consume and use PowerShell cmdlets, so you can build your own dashboards, management tools, widgets, reports, or whatever you want.
I also like that a lot of vendors are building their own in-box management tools based on their products' cmdlets. That forces them to put their money where their mouths are. Gone are the days when the built-in management tools had all these great functions and features while their scripting APIs only supported a subset of functionality.
What will be the next hype in the virtualisation area?
Hybrid clouds. (Or is that the current hype?) I like Satya Nadella's approach to this and the whole "datacenter without boundaries" thing. (This article from InfoWorld sums it up nicely.)