+David Bucci +johnny cadavid
Actually this is different from App-V in that what App-V does is essentially creates a glorified zip file that contains all the registry/files for an application and when you launch the application it uses code injection to add a filter driver which intercepts Win32 API calls to the registry and file system. If the entry exists in the sequence, it serves it up from there, if not, it passes through to the local registry.
In contrast cloud paging is essentially given away by it's name, it's like a pagefile in the cloud. So when the application tries to access a particular piece of memory (say the section of memory containing the gaussian blur function as in the video) rather than pull that via a combination of reading the DLL into memory from the local filesystem it reads it (and only it) into memory over the internet.
In the case of App-V all the data is delivered at once (it does "kind of" do some of this by using feature block 1 which basically says "what all data do you need just to launch this application" and thereby I would imagine COULD be extended to have similar functionality to this, but they have so, so, very far to go) and isolated in a container, in the case of numecent the data isn't delivered until the user actually uses it.
Which brings me to my questions/concerns. Latency being the first one. The amount of ram used by the blur function is not incredibly large, and when you don't load all the unused stuff into ram and have a "feature block one" that is 214MB given that is (depending on the version) roughly half of the size of the Photoshop CS6 install I have my doubts about the impressive look and feel. 3MB of chrome is a decent percentage of chrome. So what percentage of an app is needed for the initial stub for typical apps? (add 15% to whatever number they give you)
More importantly the same issues App-V and Thinapp and Juice and every other virtualization technology have SEEM to still apply here, namely drivers, interoperability, COM+ andshell extensions.
Case in point, if your company has warehouses or manufactures anything App-V wont work for most of your software because guess what, most of that stuff needs drivers. Adobe Acrobat needs drivers, that means most law firms are going to have a problem with this.
If you boook flights as part of your business you are likely a Sabre customer, and you are screwed.
If you use excel plugins how are those plugins going to make their way into NaaS (REALLY hate that term btw, they are pretty clearly stretching lol), for something like Office where the plugin installers smart (or dumb) enough not to check for office just dump them to C:\Program Files\Office\Excel\Plugins maybe excel would pick them up just fine, but what about the ones that dont?
What about the ones that rely on runtime integration? The same thing that kills App-V (the fact that it can't see office because office technically isn't installed) is going to kill this.
When you go to sharepoint and click "open in excel" or click a mailto link how is it going to know that the mail provider for outlook should exist or that excel is installed and this is how to launch it (oh and btw this is how to pass it the target file it should open via COM)?
How is windows supposed to know when you right click on a zip file to provide you with the context menu option to extract it?
The real problem with these semi hands off demos is that demos are what stuff like this is MADE for. Because it looks incredible, and it IS incredible. Technically speaking I respect the underlying approach these guys use FAR more, it's still "wrong" to call it virtualization but it's a hell of a lot more sophisticated than what ThinApp or App-V does.
And there are obviously a good number of use cases where this does, in fact, make tremendous sense, and I heard him use at least one of the holy IT trinity, financial, medical and education. And the reason why those guys have it so well off is because a) regulations b) control. A bank has many regulations and a LOT more at risk than the average company so the amount of control they can (and most often flat out have to in order to comply with certain laws) means they have no choice but to restrict their software/usage, and that restricted usage means a LOT fewer people needing random extensions for this or plugins for that or oddball developer software. These are also three examples of incredibly focused business fields. Fields where the overwhelming majority of the company is dedicated to a specific task.
Compare that to a company that I recently worked for where they had somewhere around 8 distinct "disciplines" and the software count to go with it.
Anyway the end of the day here the value in App-V and things like Numecent are in removing support costs where application delivery is concerned. And you do that in no small part by headcount, which is fair enough, but harder than it seems, because if I can get 80% of my software virtualized (and for most enterprises, that is VERY lofty and few outside of the holy trinity reach that) then I still need testers and packagers for the other 20% and I still need to support that 20%, though in fairness when I say support I largely mean helpdesk calls, which in a decent organization most of those calls center less around failed installs and more around problems with the software itself, but it's something, and I suppose if you have 6 packagers and 4 testers this would help you reduce SOME headcount, but in a business with one or two, that remaining 20% of applications means you are now paying a subscription fee to numecent AND headcount for the other 20%.
Then there is the issue of in-house B2B apps, but that is a whole other thing (and one that is just typically crappy no matter what you do because good developers are hard to come by, and ones that have an interest in cleanly installable software even more so lol).
And since this is a WALL of freakin' text I'll just go ahead and say also that twenty years from now this may absolutely be the way 99% of software is used, I really hope so. But considering how much ANCIENT software and how utterly resistant ISV's have been to improve the situation. It's been how long since Windows Installer was introduced and it only finally about 2 years ago started living up to half the promise that was expected when it was released, and there is still a sea of poorly written installers and nullsoft installers and people who package new software in old versions of Wise and InstallShield because they don't want to pay to upgrade and people who have locks on their market so frankly don't give a crap.
If you JUST use an app that doesn't interact with any other apps and doesn't use drivers and doesn't use shell extensions and doesn't use COM+ then yes, this is AWESOME...but so is App-V and you don't have to pay a monthly fee per app to use it.