Posted by Ron Oglesby on Fri, Feb 05, 2010 @ 11:41 AM
There are many truths in computing. So many that we don’t even think about them as “truths”, they’re just there all the time, and we just teach new IT folks to accept them. But every now and then something throws one of our truths under the bus.
VDI has tossed the truth of “Disk is Cheap, Memory is Expensive” under the bus for persistent desktops.
Obviously if you compare GB to GB, disk is less expensive than memory. But if you take a deeper look at VDI and break down the hardware cost of supplying a virtual desktop, you will see that disk is much more expensive than memory as a line item per VM.
How do I back this theory up? Well let’s take a VDI environment that is expected to perform at a pretty high level. For a standard VM in this environment you will probably assign one GB of memory, maybe two. On the outside you may assign four GB if you are really looking for performance. So let’s take those three numbers as an example and look at memory cost.
If I look at HP and Dell’s websites (and I just did) and take the “work horse” servers of their lines – the HP DL380 and Dell R710 – you can come up with a pretty simple cost model.
For the HP server it costs $1855 for each 32GB upgrade. Simple math shows this comes out to about $57 per GB.
For the Dell PowerEdge R710, it costs $400 to upgrade from the standard 8GB of cheaper memory to 8GB of RDIMMs, and another $1400 to upgrade to 32GB of RDIMMS. So again, we are in the $1800 range for 32GB of memory, or $56-57 per GB.
Assuming our three VDI memory configurations we wind up with the price per VM for memory at:
- 1 GB per VM - $56.25
- 2 GB per VM - $112.50
- 4 GB per VM - $225.00
Now, let’s look at disk. For this comparison I will exclude the cost of fibre connections, SAN switches, etc, as these may be required by your virtual infrastructure anyway. To even lean towards making it unfair, so you don’t think I am stacking the deck to prove a point, I will start by only looking at the cost of LOCAL disk.
With an HP Server you can buy a 72GB, 15k drive for about $350. Assuming you configure them in a RAID5, with 6 total drives you would have about 360GB of usable space. Assuming you leave some room for swap files and what not (15% of the available space) you wind up with approximately 305GB of truly usable space. So for $2100 in disks (6 drives x $350) you get 305GB of storage for your VDI virtual disks.
That comes out to about:
- $6.88 per GB, or
- $137 for a 20GB VMDK, or
- $206 for a 30GB VMDK!
And remember this is LOCAL 15K SAS drives. Not solid state, not high-end disk, no tray cost, just local SAS disk. Once you centralize this storage and include all SAN costs it is not unreasonable to see these prices double and is quite normal to see costs per GB in the $18-$20/GB range. So let’s compare our disk costs using local storage vs more expensive SAN storage.
- Local storage using 25GB @ $6.88/GB = $172
- SAN storage using 25GB @ $18/GB = $450
As you can see, even using the lower-cost local storage, a VDI desktop using 2GB of memory and 25GB costs you $112.50 for memory and $172 for storage. Moving that storage to the SAN creates an environment where memory cost would still be $112.50 but you are now paying $450 for SAN storage!
Thus another paradigm shift we are seeing caused by desktop virtualization: Disk is expensive. Memory is cheap!
Posted by Ron Oglesby on Thu, Jan 28, 2010 @ 03:05 PM
Today we were having a discussion with a customer about virtual desktops and what defines success in a VDI implementation. It came down to this: basically a complete VDI solution must allow for simple and scalable management of the environment, personalization of the desktops deployed, and minimal use of storage to keep the cost low enough to be palatable.
At this point my brain flashed to that old adage about “Speed, Quality, or Price, pick two”. The idea is that in any project or product implementation, these are your tradeoffs.
Sure you can have a high quality product to market really fast, but it will cost you. Or you can create a product really fast for very little money, but you wouldn’t want to use it to fly airplanes. You get the point…
And instantly I realized that we have the same type of triangle today with VDI, except it’s “Management, Personalization, or Storage, pick two”.
Much like the original product triangle, VDI today suffers from the “pick two” issue. If you want significant storage savings to bring cost down, you will most likely lose persistent personalization of the virtual machines. If you want customization and personalization of every desktop, you lose some of that nirvana of “single image management” or maybe the storage efficiencies. If you have to have a small storage footprint and it has to be easy to manage, then persistent personalization that “sticks” through upgrades and changes to the centralized images is often lost.
When I look at the VDI landscape and talk to customers attempting to realize the promise of VDI, I continually hear about them having to change their design to trade off one of the sides of the VDI Triangle.
At Unidesk we have figured out a way to provide all three sides of the VDI Triangle. Our solution allows for management of the OS and applications in a very simple fashion, persistent personalization of the virtual machines to give every user a unique VM desktop, while still providing the storage savings that are only possible today with thin provisioned or “linked clone” technology.
So does VDI success depend on the triangle I describe? Is it a triangle or am I completely off base and it is really a pentagon or octagon? Would love to hear your thoughts/experience along these lines.
-Ron Oglesby
Unidesk Chief Solution Architect
Posted by Nicole Reineke on Fri, Jan 22, 2010 @ 09:27 AM
Outstanding sleuthing - the leaks are true! Unidesk has closed $12 million in Series B funding. Erin Kutz of xconomy broke the Unidesk news yesterday based on our SEC filing, and it was quickly re-tweeted throughout the industry.
Our formal announcement will be coming soon, but we wanted to let you know that we are thrilled to have such strong financial backing. Don’s quote sums it up for me:
“It’s never easy to raise money, especially in this tough economic climate,” said Don Bulens, President and CEO, Unidesk. “It’s rewarding that our investors have such confidence in our opportunity, vision, and team. Their strong support enabled us to execute this financing efficiently and stay focused on creating enduring value for our customers and partners. We look forward to rewarding their commitment by making Unidesk the de facto management standard for desktop virtualization.”
-Nicole Reineke
Director of Product Management
Posted by Nicole Reineke on Sat, Jan 09, 2010 @ 05:37 PM
Several times a year, the Unidesk employees pull together to support our favorite charities. Given our affinity for building and design, we decided that for this holiday season's event, we would all work together to create a Lego village, voting on the most creative and best executed entry. The winner got to pick their favorite charity to receive the donation.
Each employee was supplied with a base and a set of similar Legos, with only 1 rule: to bring our city together, you need to have a road running in a straight line through your base. Many of us joined forces and pieces and worked in teams, while others recruited their children to help. Together, we created 15 incredibly cool entries.
The winner was a cityscape depiction of our software, with 
buildings, cars, bikes, and people representing the Unidesk virtual desktop management and delivery platform, creatively executed by our QA department. Way to go team! Stop motion animation videos soon to follow ;) Following a close second was a scene straight from the movie “Fear and Loathing in Las Vegas,” while third and fourth depict beautiful island beaches. Other entries included Jurassic Park, killer robots, a resort with pool, windmills, green houses, and a dog park.
Check out some of the Twitterpics of our first and second place winners:
All of the entries were outstanding, creative, and just plain fun. I am already planning next year’s entry!
-Nicole Reineke
Director, Product Management
Posted by Tom Rose on Thu, Dec 31, 2009 @ 08:35 AM
As we welcome in 2010, we'd like to take a few minutes to reflect back on some of Unidesk's 2009 highlights (in no particular chronological order):
Since everyone else is making 2010 predictions about VDI and the desktop virtualization market, we'll simply say that the new year is sure to be an exciting one, with Unidesk officially emerging from stealth and our first customers sharing their Unidesk experiences.
A big thanks to our community for your feedback and support throughout 2009, and Happy Holidays and Happy New Year to all!
-Tom Rose
Unidesk CMO
Posted by Nicole Reineke on Thu, Dec 24, 2009 @ 11:41 AM
Every IT Planner at HQ liked VDI a lot.
But the Administrators, who lived in the Data Center, did not.
The Admins hated VDI, the whole VDI season.
Please don’t ask why, no one quite knows the reason.
It could be, perhaps, that its storage was too fat.
It could be that the admins couldn’t upgrade that
Application, or hot fix, or black Tuesday patch,
Without sending the users’ data down the trash hatch.
But whatever the reason
Productivity, administration, storage, disconnect,
It was an impossible mix for anyone to expect.
“And they're planning FY twenty ten,” the admins snarled with a sneer.
“They’ll demand VDI, it's practically here.”
So they stood with their tired eyes, and stomachs quite queasy,
“We must find some way to make VDI easy.”
And then they got an idea.
A clever idea.
The admins got a wonderfully clever idea!
“With this platform installed, I can make it all work!
From patching, to storage, to personalization for the jerk.”
The network won't go down when they try to log in,
And to fix them is easy: my Level 1 Service Desk can win!”
Energized and happy, the admins gleefully stated,
“I won't drop VDI, I’ll use Unidesk’s new Beta!”
(OK, you find something that rhymes with Beta. ;-)
What? Nicole being cheesy? Never! Have a wonderful holiday season everyone!
-Nicole Reineke
Director of Product Management
Posted by Ron Oglesby on Fri, Dec 11, 2009 @ 09:00 AM
Lately I have been reading a lot about user installed applications and layering in desktop environments. Great articles on these topics include Brian Madden’s A technical explanation of why the whole “layering” / shared image thing is so difficult, Gabe Knuth’s What is layering and why does it matter? and Andrew Wood's User Installed Applications – Dream or Nightmare?
Some separate layering and user installed applications into different categories. The first is a solution for many existing issues, and the other is simply a request from the user that causes more problems for IT. I personally think these two concepts are married at the hip and that one is a requirement and the other a potential solution to that requirement.
Part of the problem (as I see it) is that corporate desktop management teams are ignoring one (user installed apps), while starting to think about the other’s benefits (layering). They're focusing on layering concepts as a way to ’manage’ the desktop while allowing some personalization/unique configurations, a problem that has plagued IT since the early days of server-based computing.
But you can't maintain control of the desktops as if it's still 2000, when the idea of mobile users and laptops in every cube was a pipe dream. As I see it, layering is an opportunity to give mobile and in-office users more control of their environment, while still enabling IT to effectively support them, and at the same time solve the profile and patch problems.
Giving the end user “real” control means allowing them to install the dreaded “user installed applications”… but we need to get over it. The idea of installing an application ourselves is normal. We (IT people) do it all the time and think nothing of it. But we still hold users in contempt when they ask that they be granted this godlike ability. Why?
In corporate IT the world over, the ‘why’ is simple: users can break their desktops if they install something. It used to be that users weren’t savvy enough to install apps and often made inadvertent changes to their systems when they did. But users are savvy, often way savvier than we give them credit for. Try to find someone under 35 (a large part of our workforce) who hasn’t used a computer for almost a decade - they probably own at least 1 or 2 at home.
Are these people as good at remediating install problems as IT? No, of course not. But they are constantly running into times when they need to install items to get actual work done (think Webex, gotomeeting.com, ActiveX controls, a file reader for a specific file type, etc).
With the advent of the corporate laptop, they are now encouraged to carry their PCs to every meeting, on the train, to their homes, on airplanes, and even on vacation! And we would expect them to do this without installing iTunes?
So, are user installed applications important in the corporate desktop world? Yes. And not just for recreational purposes.
While this thought is enough to make desktop managers and helpdesk staff wake up screaming in the middle of the night, it’s time to look for the solution and stop ignoring the demands of today’s users. That was the whole starting argument for a lot of VDI admins saying “I need to give users administrator level privileges.” Yet you still see locked down desktops, profile tools, folder redirection… starting to look like early Terminal Server solutions huh?
The idea of user installed applications is not new. But the idea of embracing them in a corporate environment is. Layering the user’s desktop environment into distinct components allows for unique configurations, application deployment, patching and even user installed applications. This is the future. How this layering is done in the end is still up for debate, but it will be accomplished.
So the question is do you want to continue chasing the policies and systems of 2000? Or will you move into 2010 looking for greater user satisfaction, better productivity, more flexibility and, yes, user installed applications.
-Ron Oglesby
Unidesk Chief Solution Architect
Posted by Tom Rose on Thu, Dec 10, 2009 @ 11:30 AM
Industry analysts are continuing to track Unidesk in their desktop virtualization and PC configuration management research:
While we're excited and appreciative that the experts have taken notice of our technology, we're even more excited that we're moving beyond demo. I've been running my Unidesk layered desktop for my daily work for quite awhile now, and we're about to start our first customer installations. Stay tuned for details, or register for Beta to experience Unidesk for yourself.
-Tom Rose
Unidesk Chief Marketing Officer
Posted by Nicole Reineke on Fri, Dec 04, 2009 @ 07:19 AM
This month, Pragmatic Marketing published my article on the value of optimizing and creating meaningful relationships to enhance search. While I wrote this from the perspective of making it easier for people to find you for marketing purposes, it started based on research I was doing for our own product development.
Shocking, I know. Real thought does go into marketing. ;-)
But the marketing search concepts I wrote about are simply another practical application of good user interface design principles, based on understanding how people find, filter, group, and use information.
So how does this relate to Unidesk? Our product design problems were fairly complicated. Our virtual desktop management platform has users at opposite ends of the experience spectrum: from entry-level Service Desk administrators who need to manage thousands of desktops (and who may even be working for an outsourcer) to savvy IT administrators who need to perform more complex operations like OS migration, application upgrades and distribution of disparate software.
When we were doing usability design internally, we made a conscious decision to ignore the “standard” approach used by most desktop administration tools (many of which were built decades ago) and design the product from the ground up to do the most good for our IT users.
Why does this matter? If we were to base our product on more traditional applications, we might have provided the ability to search on username and desktop name. Maybe forced the user to browse a tree (let’s hope your LDAP guys have it together, or who knows where the users live.) Or some other arcane methods of hunt and peck. That can be awfully cumbersome to learn for those (sometimes high-turnover) entry-level administrators accustomed to information at their fingertips.
Instead, we approached the problem from a different perspective, did testing and mock-ups, and came to the conclusion that Apple and Google have it right: let the IT user tell us what they are thinking, whether it is about the OS, application, desktop, user or even OU (or many other options). Display the ‘results’, and enable them to group, tailor, and act on the results on the fly.
With this approach, the level 1s can find the desktop or group of desktops that they want to work on quickly and easily, and administrators can efficiently examine the topography from the Operating System or Application view, and then perform actions against the grouped items. All from just a “simple search” that exploits our knowledge of relationships.
So how does this reduce OpEx? By presenting inherently complex information and relationships in a simple way, common IT administration tasks like finding and fixing or updating desktops – which often require hours of a level 3 administrator’s time - can now be quickly completed by an entry-level administrator without risking data loss or end-user downtime.
Customers and VARs are telling us that ease of implementation and management – of which simple search is just one example – is one of the things that distinguishes Unidesk from competitive solutions. Are they right? Sign up for a demo, or register for our Beta and let me know.
- Nicole Reineke, Director of Product Management
@NicoleReineke
Posted by Tom Rose on Mon, Nov 23, 2009 @ 11:00 AM
Our Beta participant list is growing exponentially, but it isn’t the numbers that are most interesting - it’s the diversity. In the beginning, we heard from the usual VDI early adopters: financial services, higher education, government, and health care. Now we’re seeing greater variety: manufacturing, electronics, human resources, consulting services, Desktop as a Service (DaaS) providers, … wait, did I say DaaS providers?
Providers of DaaS give customers “…a Windows desktop provided by an outside company, rather than delivered by the internal IT department" (see Brian Madden's recent article “What is a Desktop Cloud?). Those offering DaaS range from venture-backed startups targeting small businesses (think <250 desktops) - to VARs looking to expand their services offerings - to global companies planning to make DaaS a new profit center. Regardless of their size or target market, we’re finding they all struggle with the same challenges – how to make DaaS affordable to operate, and attractive to customers.
In the words of one DaaS vendor:
“It’s all about margin and economies of scale. If I can simplify desktop management so I can manage hundreds of customers as easily as managing a few, then my only incremental cost is adding more blades and a little more storage on the back end to accommodate new customers. My margin stays basically the same, and my business grows proportionally with my customer base.”
The DaaS vendors coming to us have been trying to use VDI as the foundation for their hosted service, but they’re hitting the same implementation obstacles as enterprise IT, where VDI can be compared to a private desktop cloud. Except with DaaS, these problems are magnified because they span multiple customers and thousands of desktops:
- Windows image management: “We were hoping to have 1 gold image to patch and manage, so we’ve been testing Linked Clones. But there are some real storage cost and performance issues around the growth of redo logs. Combine this with the loss of user changes whenever we modify the customer’s gold, and we’ve been forced to create persistent desktops for every user. This is just blowing our storage cost model out of the water.”
- Application packaging and delivery: “Application virtualization is something we’re trying out, but not all apps virtualize, and we find we have to understand a lot more about our customers’ business and about what apps need to communicate. It’s adding a lot of complexity just to solve the app delivery problem, and that means more staff on our side, which starts to impact our margin.”
- User personalization: “It’s hard taking over desktops for a SMB, where they don’t have a lot of IT discipline, and where users are accustomed to doing whatever they want on their PCs. They expect the same desktop experience with DaaS, and they don’t have a good reaction to being locked down. We’ve looked at profile management tools to help, but they don’t catch everything the user does, and they add more cost that we either have to eat, or pass along to our customers.”
Unidesk is an enabler for DaaS providers. Unidesk’s ability to solve all of the common VDI problems – image management, application delivery, user personalization, storage – with one virtual desktop management platform means DaaS providers can take on more customers with the same desktop administration and support staff, and less capital outlay for storage. And by letting users have a rich, customized desktop that is still easy to provision, patch, secure, and repair, the customer acceptance issues go away.
We’re excited to have so many DaaS vendors signing up for the Unidesk Beta program, and we look forward to working with them to bring the benefits of Unidesk to customers we couldn’t otherwise reach.
- Tom Rose
Unidesk CMO