Do we all agree that ESX is the only hypervisor to consider for enterprise size VDI deployments?

If so why?

If not, why not?

I do not agree at all. The hypervisor has turned into a commodity item. We have ESX, XenServer, Hyper-V and Paralells to name a few.
So how do you choose the Hypervisor for VDI.
1) What VDI solution are you using? Are you using Citrix, VMware, Quest? It may make sense to bundle the hypervisor into the product you purchase, or if it is VMware, then you have to choose ESX.
2) Do a workload comparison. If Hypervisor A can handle 45 workstations when users are using them and Hypervisor B can handle 50, then at around the 1000 concurrent user mark you will notice a cost saving as you have purchased 1 less server which means less infrastructure, less managament cost, less power etc.
3) Compare cost. Do you really need a fully redundant, highly available desktop infrastructure? Would it make sense not to go with a HA/DRS solution and go for something "free"? Conversly do you have SLA's on your desktops or bringing them in because of VDI so you have to incorporate this functionality?

Just some thoughts from my side of the ocean :)

Seems to me that as much as we might love to have a homogenous world with one hypervisor to manage, the realities of business will break that rapidly. Acquisitions bring mixed solutions; price / value equations drive multiple vendors (Xen vs. ESX, for example); Vendor support requirements force platforms (MS SQL -> HyperV, Oracle -> VI).

Additionally, there is a lot of innovation going on right now in the "infrastructure" layer (hypervisors in particular). This is especially true with the new client type 1 hypervisors.

This implies that when building an architecture for virtual desktops, it will be important to make sure that all software that lives "above" the hypervisor, and is agnostic to the hypervisor - supporting all of them. Then we can pick the hardware layer, and the hypervisor/infrastructure layer based on best-of-breed/price/etc and not stuck with vendor lock-in.

I'd have to disagree (but then I am slightly bias!)

I work for Parallels, and we're currently having a lot of joy with our joint Parallels/Quest solution, whereby the infrastructure layer is provided by our Virtuozzo Containers platform. We've recently won some large deals over VMware/Citrix, and have even been selected as the VDI platform in some die-hard ESX Server houses, and there are 2 primary reasons behind this :

1) Containers are much, much leaner than a VM. They use a fraction of the memory and disk as they are not full OS's (even thought they behave like on), meaning we typically get 2-3x the number of users on a hardware node than our hypervisor cousins. This clearly leads to some pretty serious economies of scale in areas such as hardware, maintenance contracts, racking, power, cooling, UPS, cabling, networking etc etc.

2) The Microsoft VECD licenses don't apply to Parallels. As a fleeting overview for those unfamiliar with us, we 'slice' a single server OS (windows or Linux) into isolation workload partitions, much like Solaris Containers or IBM wPARS. In a VDI scenario, we are therefore partitioning a single Windows Server 2003/2008 OS into multiple user spaces, into which we apply templates into to appear to the user as their trusty desktop OS.

From a Microsoft licensing perspective, this means that the Windows Datacentre licensing model (which allows for an unlimited number of virtual guests) covers the OS licenses for every user on the box - which as per point 1) is a very large number of users indeed. Microsoft have also stipulated that a TS CAL is required per user, which when added to the DC license equates to a much smaller number than the VECD cost. not to mention a licensing scheme which is a great deal simpler...

Now there are caveats to the Parallels offering... firstly, the architecture demands standardisation per host - so no mixing and matching XP/Win7 guests on the same host (however, existing in the same 'infrastructure' is fine). Also, as we are delivering a 'Server' OS kernel there might need to be extra considerations around application compatibility (which can be solved with application virtualisation/streaming).

But all in, I strongly believe the Parallels VDI platform serves the primary requirements of an enterprise VDI deployment, which in my opinion is the following 3 criteria

1) it must be easier to manage than a traditional 'fat' deployment

2) it has to improve the user's productivity and flexibility

3) it has to be attractive from both a capex and open perspective

I'd suggest that when correctly planned and configured, all of the platforms will meet criteria 1 & 2 to some degree - however the architecture of Parallels VDI means we are in a class of our own on number 3.

I hope you've all found my ramblings useful, and I'm happy to discuss/debate in more detail if anyone is game!

Regards,

Scott Holmes

Solutions Consultant - UK Enterprise Division

Parallels

Hi, I'm also Bais but have to agree with Scott.

With ESX's poor scalibility and hi storage requirements the ROI offen becomes nearly impossible.when talking VDI, and lets face it if you use View Composer the ROI stays the same, you are just giving VMware the storage vendors money in extra license costs. (same goes with Citrix and their provisioning server).

Matt

The storage element is key. Next generation SAN (virtualized architecture) with Thin Provisioning and reservationless copy on write technologies (such as 3PAR) work well with VDI, and much improves the TCO / ROI and greatly simplifies management.

- http://www.3par.com/solutions/utility_computing/vmware_vdi.html

Mark

i have to disagree, the reason ESX has so much traction in the VDI space is due to its success in the server space and companies are familiar with the product. I am not saying that the Vmware solution is poor but people tend to stick with what they know and are familar with. (no training costs etc)

As far as I am concerned when designing a solution I am more focused on the end user experience, the flexibility of the connection broker, the "solutions" storage and networking requirements etc

Then I look at the hypervisor from a density/scalablilty/reliability perspective.

I agree w/Dave in terms of market traction. I've met with tons of customers evaluating and deploying VDI, it the vast majority are on ESX today for the very reasons he mentions. Interestingly, a lot seem to be choosing Leostream or XenDesktop for their broker over View.

Comment above from "Anonymous" was actually from me! For some reason I got logged out of VDI.com, and couldn't login again on Safari even though I shut it down and reloaded it. Seems to be OK w/Firefox...?

Hi Matt, I wonder if you have any numbers to support your post? I am very interested in cost comparison and was looking for the same kind of information.

Is this what you are seeing Dave? what sort of percentage of VDI projects are you seeing go the way of VMware?

I ask because that isn't what I'm experiencing... whilst a lot of existing VMware customers first instinct was to say "well the server stuff worked a charm, let's try VDI", they quickly got turned off the idea when the total costs of VECD, storage, hardware etc was taken into account - never mind disappointments with protocol etc - hence why they came knocking on our door, Citrix's door - and now Microsoft's door. Not to say that no-one is going VMware, but people's immediate instincts to go that way are being tempered by experience (not to mention critical journalistic pieces that have surfaced recently! )

Our Virtualization Consulting company was looking for independent VDI solutions review for one of our large customers. I believe we had reviewed Parallels and even met with someone at our Montreal office. Although I was not part of that discussion hence cannot really comment on the outcome.

Scott, could you tell if Parallels deployed its VDI solution at any customer in Canada. I'd like to hear success stories, if you could share. Even a feature/cost matrix comparison would be infromative.

Cheers,

Eugene

Hi Eugene,

I've asked my colleagues in Canada if they can submit case studies or reference sites, and in the meantime if you wanted to gauge the potential TCO savings you can visit our VDI portal site where we have a comparison calculator tool:

http://www.vdi-portal.com/en/vdi-deployment/vdi-calculator/

This is an EMEA thing so only shows in £ and €, but should give you the general idea of the areas of cost reduction. I'll be back soon with the other docs!

If you asked me 9 months ago, I would say yes, it is pretty much the only show in town. ESX is a great product and I like it a lot.

But today, I would not be so sure. I have am now wondering why people would buy VMWare when there are very credible and free alternatives in XenServer and (now) HyperV, which are deployed in production environments for some decent sized companies, and have earned their stripes.

When you cut through the marketing and sales hyperbole, many customers are now considering what they get with VSphere that justifies the additional cost. Some people look on it as "Virtualisation Taxation".

I have used and would recommend XenServer 5.5, and I like what I see in HyperV, but have yet to prove it in anger. ESX, still like it, but why should I use it today?

Maybe people still buy ESX / vSphere because hypervisor price is not main aim?

You know, I still can't understand why so much people are so worried about hypervisor cost? Why can't you worry about whole system cost including number of physical servers, network ports, disks in SAN and guest OS licenses?
Hypervisor price is nothing, maybe a couple of percents comparing to whole solution price.

So, you can make your choice based on
1) politics: all our stuff should be from
2) what your employees can: if prices for solution from different vendors are not really different - choose solution you already have trained people for
3) technology: choose solution with most advanced features
etc

But do not make your choice based on hypervisor price, that's mistake.

Matt I hear ya old buddy on your thoughts on cost but as it stands end-users don't purchase VMware technologies based on cost. Most of the time its performance, features and functionality. Putting cost aside can you give us an understanding of where Parallels stands on those 3 area's? Especially performance. From articles I've read its usual that VMWare wins on scaling multiple smaller workloads whilst Hyper-V and Xenserver win on less heavy workloads. Parallels is never in the independent assessment. If you have some examples where Parallels is also evaluated please let me know. Also you say VMware doesn't scale but in terms of VDI with VMware view you can scale to 10 of thousands of VMs. How does Parallels scale in a VDI environment?

As you know I'm familiar VMware and Citrix but not Parallels so I'd like to know where you are at now a days.

Matt the last post was mine - forgot to login..