[THIN] Re: A definitive registry information source?
- From: "Jim Kenzig" <jkenzig@xxxxxxxxx>
- To: thin@xxxxxxxxxxxxx
- Date: Fri, 9 May 2008 12:37:39 -0400
I've never disputed the fact that VDI is a great idea and a useful
technology. I just can't bring myself to see how it can be called a cost
savings to any corporation attempting to use. I'm not going to be able to
convince a board member that we need to make a million dollar investment
even based on the great points you make below. : ) It's real desktops for
me for the foreseeable future. Sigh.
Jim
On Fri, May 9, 2008 at 3:10 AM, Rick Mack <ulrich.mack@xxxxxxxxx> wrote:
> Hi Jim,
>
> I used to be a terminal server (TS) bigot too because I could get almost
> any application going on TS. I'd spend days and nights finding out how
> applications worked, and how to modify the application or environment to
> make it happy. But not everyone has the skill, experience, time, tenacity or
> stupidity to do that. I did it because it was fun and I've learned an
> incredible amount of stuff. Other people are sensible and have a real life
> :-)
>
> Things are a lot easier than they used to be. Application verifiers, the
> Microsoft Standard user analyser, the new TS application compatibility
> analyzer and application isolation and streaming have made the whole process
> of porting applications to TS a lot easier.
>
> But it's still not as easy as installing an application on a Windows XP
> desktop and unfortunately the major skill set in a lot of
> organizations resides in the desktop management team.
>
> You know them, they're the people who reboot a TS server with 50 users
> logged on because the spooler has stopped. They often just can't or won't
> come to grips with TS and actively hate it. They're the ones that have made
> Citrix a dirty word in a lot of organizations.
>
> Anyway, lets get down to the real stuff.
>
> Server based computing (SBC) is about delivering a remote GUI to users
> while running the users' "virtual" desktops or applications in the data
> center. Remote doesn't necessarity mean far away, just that the desktop or
> applications aren't running on the device in front of the end user.
>
> SBC coupled with thin clients is widely acknowledged as a brilliant way to
> keep desktop management costs in check, and to get away from the 2-3 yearly
> desktop refresh cycle.
>
> So if that's such a good idea, why do so many people hate TS?
>
> Because it's different, and it's often just too hard.
>
> It needs different skill sets, more experience and a much better
> understanding of how everything hangs together. The people who do TS well
> are the IT equivalent of rocket scientists.
>
> SBC used to be solely about TS, but not any more. We've now got all sorts
> of ways to give users remote access to their virtual desktops and
> applications. We have user session isolation (TS), operating system
> partitioning (Virtuozzo), virtual machines (VMware, Virtual Iron, Xen,
> Hyper-V etc) and even blade PCs (DDI) that weakly qualify under the SBC
> blanket because they're racked machines running in a data center. That's
> ignoring stuff like VWware ACE and Moka5 where the difference between "real"
> and virtual user sessions is just too vague.
>
> So we've got a lot of different alternatives for delivering virtual
> desktops, but which one is THE best way to do things?
>
> Let's look at our SBC options.
>
> At one extreme, if you've got a lot (200+) of interesting
> (challenging/frustrating etc) applications, then the implemetation costs you
> incur to get all those applications going on TS can be horrendous. And then
> along comes something like Cisco Presence that won't run multi-user at all.
> Oh, and the corporate application tha the CEO loves that is having problems
> and isn't supported on TS.
>
> So what do we do? Go back to fat clients for some of the users?
>
> At the other extreme DDI is brilliant from the viewpoint that you can run
> anything on a blade PC, even stuff like AutoCAD with 3D rendering etc. Any
> super-heavy resource hungry application can be accomodated and we're still
> using a thin client in front of the user. But the cost of having a thin
> client AND a blade PC for every user is pretty horrendous.
>
> Then there's VDI which is kind of a compromise. The number of user virtual
> machines you can support on a given host is a lot less than TS but VDI on
> virtual machines hosted on VMware etc will run almost any application that
> will run on a desktop. No file sharing and locking issues, no problems with
> privileges, no support issues. If you're crazy enough (which I hope no-one
> is) you can even let your users be local administrators.
>
> If I use Virtuozzo for VDI as an "intermediate" VDI solution, it's not
> quite as "good" as a windows XP virtual machine. We're still running on
> Windows Server 2003, but the user session isolation is heaps better than TS
> and the number of users I can host on a server is nearly as good as TS.
> That's "VDI" at almost the same hardware cost as TS.
>
> So we've got a lot of alternatives that are right for different reasons.
>
> Whever we quoted to do a fat client to terminal server migration project,
> we used to go nuts trying to estimate how long it would take to port all a
> customer's applications to TS, and worry like heck that we didn't cover
> everything, and that some of them wouldn't port at all. With VDI/DDI as a
> backstop, and particularly if you can publish seamless applications via VDI
> (did I mention Provision :-)), I can run the bulk of my applications and
> desktops on TS, and silo the applications that don't run or aren't supported
> on TS on my VDI machines. Or if users have to run a huge java application
> that uses several gigabytes of RAM, I can throw in a few blade PCs.
>
> In an ideal world we would use each technology as is most appropriate to
> end up with the best possible mix. Use TS for easy apps and lowest cost,
> VDI for the apps that need a single-user desktop environment, and DDI for
> the stuff that doesn't belong on a server. Imagine an SBC/thin
> client environment where everything works, where it's easy to put everything
> together and it's the most cost effective solution you can offer.
>
> Then there's the real world.
>
> Getting applications to run acceptably on TS can be a huge job with huge
> risks whereas doing the same thing on virtual machines is dead easy with
> minimal risk. Whether we like it or not there have been enough well
> publicised failed TS implementations that the perceived risk is unacceptable
> for many organisations.
>
> What a decision for a CTO, a defined high hardware cost with very
> tight implementation costs at a minimal risk, or a lower hardware cost
> with undefined, possibly huge implementation costs and an equally huge risk.
> P2V a standard SOE machine and you're nearly there with VDI, ingnoring all
> the stuff you should do to get decent scalability.
>
> And we can't forget the basic fact that for people who are used to managing
> desktops, VDI doesn't require much of a change of mindset whereas TS does.
> People can cope with the idea of virtual machines because you can manage
> them the same way you always have, and aside from the VMware etc
> infrastructure, nothing has to change very much.
>
> It's this scenario that means people will often opt for VDI instead of
> TS when they have a requirement for remote access, simplicity and the cost
> saving associated with thin clients.
>
> But it's not all bad news. VDI is still SBC, and where are the SBC
> experts, the people that know about thin clients, profile management,
> lockdown, tuning etc, all that stuff that makes VDI even more viable?
>
> That's us.
>
> We've only just begun to make a difference.
>
> regards,
>
> Rick
>
> --
> Ulrich Mack
> Quest Software
> Provision Networks Division
>
>
>
>
> regards,
>
> Rick
>
>
>
>
>
>
>
>
>
> On 5/6/08, Jim Kenzig <jkenzig@xxxxxxxxx> wrote:
>>
>> Yes but if I did VDI I would need about 75 servers versus the 8 I
>> currently use with TS, not to mention the additional VECD licenses and
>> Vista licenses that must be purchased on the workstations.
>> At 15k a piece for the servers then that is another 850,000 addittional.
>> It doesn't add up. I'd just as soon as buy them all laptops, ; )
>> Jim
>>
>>
>> On Mon, May 5, 2008 at 4:55 PM, Joe Shonk <joe.shonk@xxxxxxxxx> wrote:
>>
>>> Think of it this way Jim. 1000 users x 450 for a Presentation Server
>>> Enterprise License will cost $450,000 dollars.
>>>
>>>
>>>
>>> Joedee
>>>
>>>
>>>
>>
--
Jim Kenzig
Microsoft MVP - Terminal Services
Citrix Technology Professional
Blog: http://www.techblink.com
Other related posts:
