Implementing VDI

F

FRAGMAN BOB

Limp Gawd
Joined
Jan 25, 2005
Messages
292
Ive been tasked with planning and overseeing a VDI project for our organization. The scope of this includes replacing 90% of existing computers (~400) with zero clients and adding all server/san infrastructure (no existing infrastructure for VDI, only for server virtualization at this time) . We previously had a departmental staff of 4.5 (manager, Me, 2 tech, 1 part-time). With those staffing levels it would have been difficult to fully implement and manage a VDI project of this scope and impact to the organization.

Recently both of the techs were let go, so now we have 2.5 full time staff with one of those being a manager. I'm having significant doubts as to whether I and my manager can design, implement, and maintain VDI properly without having major issue. Improper setup of this would be crippling to our organization for a variety of reasons.

Has anyone done a project along these line help either calm my doubts or re-enforce them?
 
I would engage a few partners in your case that specialize in VDI specifically and they should have references for VDI implementations so you can refer back to validate they can do what they say they can do.

VDI is an animal. I see it all the time where customers do server virtualization/consolidation, they try to apply the same type of measurements..and try to tackle it like Server Consolidation, and they fail miserably, which leaves a bad taste in their mouth.

You should work through this INCLUDING the business you are rolling this out too, meaning, make sure that they are included during the entire process. Most IT orgs know the benefits and are sold before they start, the hard part is selling it to the end-user, and most of the time the end-user is pushed to the back of the line when they are the primary concern.

You should know this, you need to go through this in a phased approach, methodically.

1. Assessment
2. Proof of Concept
3. Pilot
4. Load Testing/Validation
5. Phased rollout
6. Monitoring

Most people fall on phase 1-2 and a lot forego 6 which is also bad. Of course this is a gross simplification of the the process, there is real meat in each step.
 
Last edited:
Those are some good basics there, I try to make sure I stick to that, but its always good to see.
I actually an employee of the company, this is not a consulting-client relationship. A lot of my concern comes from not only the rollout, but the upkeep. Is it even possible that two people can manage upkeep of VDI while all others aspects of an IT department (helpdesk, phone system, etc).
Would VDI even be more time efficient than just staying with traditional desktops?
 
Do you have an existing vSphere deployment that you can create a one host VDI cluster in and do some setup and testing with or are you starting from scratch? You mentioned you had server virtualization, but not what you had in place already.

Edit: no helpdesk sounds painful, do you currently support all 400 desktops then?
 
Those are all great questions. I can't really answer specifically what you would need, etc, as far as staffing but i will tell you that if you put 110% effort from the get go and do this methodically, the benefits to you will be enormous from a management perspective. Imagine a solution where you may only have to run updates on one Parent Image, or update to an application and have it pushed or streamed tot he desktop. Easier Profile management, especially in a "hot seat" type environment. Having the ability to rollout desktops in minutes as opposed to hours..etc.

The list goes on and on. Before I left my previous operational IT role, I had a staff of 4 that managed 3,500 users, 600 of those were VDI desktops, guess where all the Desktop support was focused.
 
We do have a vSphere deployment on 3 Dell hosts on Equallogic PS4000 for storage running ~25 Server VMs. We could probably make some storage space to play with but do not have processing or memory overhead currently.

We'd be looking at minimum 1 full Cisco UCS blade chassis and a NetApp FS3220 series for VDI.
 
Oh, so you don't currently have a large vSphere footprint yet. If you can get another host and put it in it's own cluster, you could at least do some one host testing of VDI. I probably wouldn't go in and purchase any large hardware until you've had a chance to test it and get a few users on it with their feedback and then you can tailor it from there.

Good hardware though, same as what we're speccing out to expand our vm infrastructure and add VDI into the mix. We're also looking at HP c3000 blades and Nimble Storage, Nimble has some great IOP performance which really helps for VDI, it's pretty impressive stuff.
 
While all those technologies are great, you shouldn't be looking at specs if you don't know your requirements. This is why the VDI deployment methodology is outlined the way it is. Right now you should be focused on two things, which can be done in tandem, Assessment and PoC. You should NOT be putting users on hardware at this point. You need to prove it out first. You may find that VDI may not even be a good fit based on the Assessment and PoC. That's where your concentrated effort should be.

When you make the determination from the Assessment and PoC, then you can look at specific hardware.

Take it from me, you don't wanna guess at your needs. You could be wildly overbuilt or under, both are bad.
 
ND40oz said:
Edit: no helpdesk sounds painful, do you currently support all 400 desktops then?
Click to expand...

You're talking to the helpdesk support (along with the one part time staffer), phone tech, jr network/server/storage admin, application specialist, and chief broscientist.
 
FRAGMAN BOB said:
You're talking to the helpdesk support (along with the one part time staffer), phone tech, jr network/server/storage admin, application specialist, and chief broscientist.
Click to expand...

Join the club bro. :cool:

Interesting thread, I am interested to hear what you find out and how you set things up. VDI is potentially on the cards in the mid/long term where I work, but we don't have the resources to commit to such a large change at the moment.
 
I am currently actively working on a 25 client test case and I can tell you that unlike the marketing materials from all vendors it's not all rainbows and pink unicorns.

The good news is that you don't need 4.5 people to manage a 400 client VDI once it is all set up and running, so your current staffing level of 2.5 is not actually a problem once you are up and running.

You wrote that you will be doing zero clients. There are probably some good usage cases for zero clients, but it's equally likely that your organization isn't one of them.

You absolutely have to figure out what your functional requirements are, including things like screen resolutions, multi-monitor setups, software, software licensing and license server implementation, linked clones, persistent desktops, yadda yadda yadda. For me the OpenGL 2.x requirement our apps have threw a big monkey wrench into a Microsoft based VDI plan. But that issue came out during assessment, prior to purchasing anything to make a go of it.

I am now going VMware View, and I have hardware with gfx acceleration on order (Nvidia GRID K1, yeah baby!).

Don't get sold on the NextGen Fusion-io hype, yeah the performance may be there but it's just not needed for 400 clients, you can run that off of SSDs for a fraction of a cost just fine.

You won't need a lot of hardware for a 400 client setup, but still, at that scale I'd just find a VAR, let them design it, buy the hardware from them, and be done with it. You easily have the man-power to run it, but it's not clear whether you have the time and knowledge to design it.
 
Last edited:
Vader said:
VDI is an animal. I see it all the time where customers do server virtualization/consolidation, they try to apply the same type of measurements..and try to tackle it like Server Consolidation, and they fail miserably, which leaves a bad taste in their mouth.
Click to expand...

^^^QFT

So many people, managers usually but IT people too, think since they virtualized their server stuff that desktops will be the same. No way in hell. I wish it were that simple.

You've received good advice here already. Try to start with a proof of concept environment first, then maybe work your way to a group of users. Do it slow, make sure you listen to everyone's complaints and criticisms and do it better the next time around.

I unfortunately wasn't able to persuade management a couple years ago to move in this direction mainly because the cost was a big one for us to swallow at the time, but also they're lazy. I'm not worried though really because this WILL eventually come around to us again.

My environment was based on VMware View version 4, so there are a few differences between then and what's out today. The hardest part for me though was virtualizing all of the various little specialized applications we had. Every single one of them is different and it got tough just trying to get them to even load correctly. I found that my desktops loved eating RAM like there was no tomorrow; they weren't terribly hard on CPU usage and bandwidth was never an issue. The benefits once you pull it all off...holy shit you can't put a price on that (well maybe you can but I haven't figured it up)....

New employee? Need a desktop? No problem. Spit him out a base desktop within minutes.

New patches from Microsoft this month? No problem. Update your desktop base image once; deploy to the environment after everyone clocks out.

Got a nasty virus that's a zero-day exploit? Your anti-virus software is a piece of shit and they aren't updating definitions fast enough while you're company production has come to a complete halt? Don't sit around hitting the 'update definitions' button all day or looking for various registry hacks; pull up your daily base image backup from 1-2 days ago before the virus infected the company and deploy to everyone. Bingo - virus is gone, assuming no servers or other stuff got infected.


Also, those zero-client devices are the way to go. I still have those Samsung NC240 monitors they are truly very simple to setup once, manage from a central point, and never have to worry about them again. Even today, a couple years after my own proof of concept environment, people still don't realize I think that THIS will be a real game changer in the world of IT. Desktop support techs could easily be tossed out by companies that implement this stuff hardcore.
 
Thuleman said:
I am currently actively working on a 25 client test case and I can tell you that unlike the marketing materials from all vendors it's not all rainbows and pink unicorns.

The good news is that you don't need 4.5 people to manage a 400 client VDI once it is all set up and running, so your current staffing level of 2.5 is not actually a problem once you are up and running.

You wrote that you will be doing zero clients. There are probably some good usage cases for zero clients, but it's equally likely that your organization isn't one of them.
Click to expand...
Wow, I appreciate all the info from everyone here and will keeping reading over this.

Regarding the zero client use case. I've read a reference architectures from VMWare that seem to put us in there, here is an outline of what we have now:
Type of desktop would be:
-80% of users would be mostly browser based applications, not persistent, no profile to worry about
-20% (or less)Office 13, 4 enterprise programs from sunguard, with most users only use 1 or 2 of the apps, typical office work
-about 10% of the total users have dual 17" monitors, the rest are all single 17's looking to upgrade them all to single 19s
​
 
FRAGMAN BOB said:
-about 10% of the total users have dual 17" monitors, the rest are all single 17's looking to upgrade them all to single 19s
Click to expand...

17" monitors?

37960898.jpg


Well you know your environment best. If the requirements you listed are not likely to change in the next couple of years after deployment then zero clients may just be a decent solution.
 
You must log in or register to reply here.
Back
Top