Gabe Knuth

Oracle VDI gets the Geek Week treatment...finally!

Gabe Knuth — Wed, 08 Feb 2012 05:12:00 GMT

It's been a while since we had a "Sun Ray = the Best Way" war on BrianMadden.com, but our most tenured readers will remember the small but vocal group of supporters that Sun Ray terminals and Sun/Oracle VDI have. The truth is, when it was a Sun solution, it only rarely made it to the top of our minds. There was little exposure on our part, and little marketing on Sun's part. Now that Oracle is running the show, we've seen changes on both fronts, and we finally reached out to Oracle to get a demonstration (and explanation), Geek Week style.

If you don't remember Geek Week, it's a time in March/April of 2010 when profiled a single VDI solution each day, going from bare metal at 8:00 AM all the way to provisioning desktops and user experience testing, sometimes well into the night. We had a lot of fun, but we didn't get to Oracle VDI because we let the community decide which solutions we should test. For the original Geek Week, we had VMware, Citrix, Quest, Microsoft, and Virtual Bridges.

We tried to keep the Geek Week spirit alive during our day with Oracle, and we were able to preserve some of the aspects of it. We fell a bit short on the user experience testing because our Apposite WAN emulator didn't appear to get along with the ALP protocol that Sun Ray clients use, so we just did LAN testing. The rest of the day, though, went according to plan, and really helped us to understand the Oracle platform (I think. If I get it wrong, I'm sure the fanatics will let me know in the comments, which I encourage :). We had John Renko at our offices in San Francisco driving the demo for us, answering all of our crazy questions.

With regards to the videos, we tried to be brief with them this time around, and not show the configuration screens and all. We've limited it to two videos, with the whiteboard video coming in under twenty minutes, and the user experience video coming in under ten minutes.

To the whiteboard!

Before we started, we had John explain to us exactly what we would be setting up, along with a quick walkthrough of our demo environment. The best way to get this information is to watch the video. The first part is a walkthrough, and the whiteboard explanation starts 3:40.

Basically, the Oracle VDI solution entails these components:

Solaris - The OS on which the rest of this stuff runs. Installation was easy, and while it may not be Windows, it shouldn't be a showstopper. Installing software is installing software.
VDI Core - Handles administration, pool configuration, cloning, provisioning, Sun Ray Server, brokering, and so on. It runs as a service on the Solaris box, and there would typically be many of these for redundancy/scalability. Typical environments have one VDI Core per 500 sessions. In the video, John talks a bit about density, but keep in mind those numbers are in a vacuum.
MySQL database - used as the data store
Hypervisor - They use VirtualBox, of course. They also support Hyper-V and ESX, but you lose some of the Oracle-only features like vRDP, which we'll get into later.
Storage - In our case this was on the lab box as a volume formatted with the ZFS filesystem. Typically this would be running on an Oracle Unified Storage Appliance. ZFS is required, so you're locked to Oracle for the time being.
Sun Ray thin clients - Pretty much the same zero client we've been hearing about for years. They have new form factors, all in ones, and so on, but by and large it's the same old thing.
Oracle Virtual Desktop Client - OVDC is the software client for use on PCs, Macs, iPads, and so on.
ALP Protocol - This is the remote protocol that the Oracle system uses to communicate between the VDI Core (specifically the Sun Ray Server).

The connection workflow as described in the video, works like this:

The client (Sun Ray or OVDC) makes a connection to the VDI Core via the ALP protocol.
The VDI Core, which is aware of all the sessions and hosts, communicates with the hypervisor.
The VDI Core then communicates with via the vRDP protocol to the hypervisor in the case of VirtualBox, or waits for the machine to boot and uses MS RDP to communicate with the VMs in the case of Hyper-V or ESX.
The VDI Core translates that vRDP or MS RDP data into ALP and passes it down to the client.

vRDP runs out of band, similar to SPICE, which enables you to watch machines boot. While it's based on RDP, it's not 100% RDP compliant. Still, there are things like the out-of-band hypervisor tie in, some graphics enhancements, and some multimedia redirection that they can do with vRDP and not RDP. Of course, you can also use MS RDP to connect to VMs running on VirtualBox, but that's a more traditional solution that relies on the Remote Desktop service in the VM. Oracle recently released a XenDesktop connector, as well, which uses MS RDP between the VDI core and the XenDesktop host, but not HDX.

Preparing and provisioning desktops

We're skipping some of the more mundane sections from Geek Week, and part of that is the installation of all the components. The fact of the matter is that it's not that big of a deal, and you don't need any special knowledge to pull it off, regardless of the fact that it all runs on Solaris. The process, roughly, goes:

Install Solaris
Configure ZFS volume (if not already done)
Install VDI Core
Create/import desktops
Prep desktops
Provision desktops

The interesting bits there are the preparation of the desktops and the provisioning, but even those are straightforward. For our test, we had base Windows 7 VM's already created in VirtualBox, and importing them into the VDI core was no big deal. Preparing them amounts to making the normal tweaks, and running their FastPrep, which is like all the other <adjective>Prep tools out there. It prepares the machine for cloning much faster than SysPrep because it leaves out the SID replacement. It also installs the tools package, joins it to the domain, and makes it available to the VDI core for pool configurations when it's done.

Oracle VDI supports three types of pools:

Dynamic Pools (non-persistent, everything is wiped at logoff)
Growing Pool (non-persistent OS, but with linked clones so user state persists)
Manual Pool (persistent, personal VMs)

In terms of guest OS support, Oracle VDI supports Windows from Windows 7 back to 2000, although you can install pretty much anything in VirtualBox and have it work. I've seen demos with Windows 3.1 and DOS. Linux, Solaris, Android...you name it, it all works. It can also broker connections to Terminal Servers (RDSH). The reason for this is the integration between the ALP, the VDI Core, vRDP, and VirtualBox. There's essentially an extra layer in between, and that layer gives much more flexibility than what we're used to seeing with the more typical desktop virtualization solutions.

This is the same diagram as above, just here so you don't have to scroll.

There are some limitations that this brings in, of course. Translating from native protocols to something else negates any enhancements built into the native protocol. In a way, this is sort of like how HTML5 clients use a gateway service to interact with the host via RDP (or whatever protocol the host uses), then re-encodes that into a text stream for the browser to decode and render. In this case we're not changing the data to text, but we are re-encoding it.

The reason for this architecture has nothing to do with VDI, though. ALP, or Appliance Link Protocol, has been around since the late 1990's as the protocol used by Sun Ray terminals. In fact, when you boot up a Sun Ray, the logon screen that you eventually see is actually being sent via ALP from the Sun Ray Server. In our case, the VDI Core was fulfilling the role of the Sun Ray Server, but Sun Rays are a thin client solution unto themselves. In fact, the Sun Ray Server has a built in RDP client so that it can connect terminals (via ALP) to RDP sessions running on Windows boxes...all without Oracle VDI or VirtualBox.

It seems to me that the VDI portion is more or less done because they can do it, and is more or less an amalgam of all the different things that Sun/Oracle could already do. They had the thin clients, they had the hypervisor, the hardware, the storage, and the base OS. The real power, apparently, is in the Sun Rays.

The devices themselves are fairly unremarkable (it's the management that really shines). They're a zero client, and I'll acknowledge that they are the first of the breed. Oddly enough, though, they still run a firmware. We know this because it was one of the first things we saw on the day Oracle came to visit. So, just because something is zero doesn't mean it doesn't have firmware, it just doesn't have any intelligence besides, "seek server...find server...do what server says."

What warrants a deeper look someday, though, is the Sun Ray Server component, and about how to use these devices. We saw a few gee whiz features that, unfortunately, didn't make for very good video (although we tried). Being able to group terminals and provision single or multiple desktops to them as if they were one box, for instance, is pretty neat, although probably not used en masse (this is called a Multi-Headed Group, which you can read more about in the Sun Ray Server Admin Guide). I really want to like Sun Rays, and the nerd in me is behind them, although I'm not sure, based on the user experience testing, that I can completely get behind the whole ALP + translated-other-protocol thing yet.

The User Experience

Admittedly, this test is incomplete. We were only able to test the LAN scenario because we didn't trust the way our Apposite box was performing. We believe it has something to do with the way ALP and the Apposite handle Path MTU Discovery, but were unable to come up with a solution that would allow us to reliably test the WAN scenarios. This is another thing that we'll revisit when the next version of Oracle VDI comes out, perhaps on site at Oracle using their Shunra.

We did, however, get a good LAN test, and the best way to get a feel for ALP aside from using it is to watch the video.

I will say that experience looks fine most of the time, but it doesn't "feel" as nice as some of the other protocols out there. I think this is because the system handles acceleration is pretty complex with two protocols in the mix. Any acceleration native to the connection between the Sun Ray Server (or VDI Core) and the virtual desktop has to be translated or passed through to the client. This results in a lot of behind the scenes work going on, and culminates in that not-quite-normal feel on the user side.

To look at how Oracle handles acceleration, there's a few terms to know ahead of time:

RCA - Rapidly Changing Area - The system detects RCAs at the hypervisor with vRDP, where it can automatically transcode those areas of the screen to MJPEG.
Sun Ray Windows Connector - This is basically the built-in RDP Client that the Sun Ray Server uses to talk in MS RDP to Windows. The Windows Connector can identify RDP 6 bitmap acceleration and convert that to MJPEG, which is then sent to the client as an RCA.
MS RDP - Microsoft RDP, which we talked about earlier
vRDP- Oracle's RDP implementation at the hypervisor level. As discussed before, this is based on MS RDP, but is something like RDP 6.1 compliant, not RDP 7. I could be off a version there.

This is complex stuff, so I hope I get it right. You should check the comments and look for Craig Bender's notes to make sure I didn't completely screw this up :)

Multimedia and Flash Acceleration

In general, if you choose to use vRDP as your protocol for any virtual desktop (this is between the virtual desktop and the Sun Ray Server, because the client > Sun Ray Server is always ALP) and an RCA is detected by the hypervisor, this motion is converted to a Motion JPEG and is sent directly to the client (not translated by the Sun Ray Server or VDI Core). The Sun Ray client has a MJPEG decoder built into it, which makes this more efficient than trying to remote the moving areas. On the user side, it's noticeable, but not awful.

If you choose to use MS RDP on Windows XP/2003 (so, RDP v5.2), Flash is handled by a feature called SunFlash when the Flash element is accessed in IE. If it's accessed through anything else, it will simply be remoted like anything else. For VC-1 (Windows Media's codec), H.264, and MPEG2 videos played in Windows Media Player, the video data is redirected to the client where it is decoded locally. The client can be a Sun Ray or the OVDC software client.

If you choose to use MS RDP on Windows 7/2008 R2 (so, RDP 7), it leverages Microsoft's RCA-like solution (which they call "enhanced bitmap acceleration." You can read more here) to know what to accelerate. Essentially, if the Sun Ray Server sees RDP accelerating video, it grabs it, converts it to MJPEG, and sends it to the client.

I asked why the fancy codec detection didn't exist in Windows 7/2008 R2, and the answer was "When's the last time you used WMP?" Touché. They said this acceleration was written before Flash was the dominant video delivery method. Oracle is also aware that this means that the experience isn't optimal, which you can see in the video, and that while they can't say what's coming, it's safe to assume something is coming :)

Wrap-up

The protocol transcoding, while complex and a bit awkward feeling right now, is still adequate for business use today. I'm hopeful that there will be some marked improvements in the future when it comes to acceleration because the competition is getting incredibly stiff. RemoteFX, HDX, PCoIP, EOP, etc... feel a bit further ahead of the pack when you're actually using them. As this technology segment grows, we have a lower threshold of pain when it comes to the user experience. Based on the many conversations I've had with Oracle in the past few weeks, I think they get that, so I'm optimistic about future changes and simplifications.

Admittedly, I didn't get to see it in a WAN scenario, but Oracle seems to be pretty proud of the way it performs. When we get a chance to do it, we'll be sure to add to the user experience video shot here. I really want to get a look the Sun Ray Server next time we connect, even more than the Oracle VDI pieces, I think.

It would be interesting to see a comparison of the backend technology that makes up Oracle VDI, so if anyone reading this feels like doing that, let me know what you come up with (or have come up with, if you've done it already). I'd be curious to compare storage performance and hypervisor performance with others. Of course, I'd also love to see Benny Tritsch and Shawn Bass add this to their already amazing session called "VDI Remoting Protocols Turned Inside Out," which, if you haven't seen it, is awesome (and you can watch it right now!)

Oracle wanted to show pre-release code, which we wouldn't have been able to talk about, so we decided to show what is available today. We'll be catching up with Oracle again when they release the next version, so if you have any questions or things you'd like to see, let us know in the comments. We'll also plan on doing a more hands-on review to go into some more options.

If you want to take a look yourself, you can probably stand up a test environment in a day or so with no Solaris knowledge whatsoever. If you happen to have Windows 7 VMs in VirtualBox already, you can get it set up even faster. If you're already an Oracle customer, you can download version 3.3.2 today by visiting http://support.oracle.com and downloading the appropriate patch. If you're not a customer, you can download version 3.3.0 by visiting http://edelivery.oracle.com.

For more information and to follow Oracle VDI and Sun Ray solutions, follow @ORCL_Virtualize on twitter and check out their Virtualization and ThinkThin blogs.

OnLive loose ends: Are they using VMware? Dedicated hardware? Custom licensing? Plus, are they on Microsoft's acquisition radar?

Gabe Knuth — Fri, 03 Feb 2012 05:01:00 GMT

Last week, I posted an article that breaks down my thoughts on OnLive Desktop, the desktop virtualization solution offered by cloud gaming provider OnLive. In the article and in the comments, we tried to figure out exactly how OnLive can deploy Windows 7 Enterprise desktops to users, and how that licensing is handled by Microsoft.

Without rehashing the entire discussion, OnLive was mum on how they've managed to license both Windows and Office in a way that enables them to deliver them to users either for free or for $9.99/mo. We went on to wonder if they could possibly have a situation where they were using valid VDA or SA licenses in a virtualization model, if they were using blade PC's, or if they had some custom arrangement with Microsoft.

Are they using VMware?

Then, over the weekend, a reader by the name of "lodani" (also on twitter as @LoDani) sent along this screen shot that shows elements of VMware exist in the OnLive file structure.

After digging around myself, I can't find anything other than empty subfolders underneath the C:\Program Files\VMware folder, so I'm guessing that the image was probably built on VMware Workstation, but then the bits were stripped out when it was deployed as a master image. In fact, with the slight amount of cruising the filesystem that I was able to do, I can't find any traces of other virtualization tools packages, so I'm thinking run-of-the-mill virtualization is out of the picture.

The custom hardware option seems to be more likely

What I did find while poking around, however, was a folder called Microsoft XNA. XNA is the core component of the Microsoft Game Studio, which aims to eliminate boilerplate code from Windows-based games to make them more efficient. To me, this leads me to believe that they may in fact be using blade PC's or other dedicated physical hardware to deliver these desktops. As mentioned in the other article, this also makes sense from a gaming perspective where the hardware requirements of these games are quite possibly more than virtualization can handle. It also jives with the fact that OnLive can leverage otherwise unused hardware during the day when gaming subsides, then ramp up the gaming again during the evening when the need peaks.

Keep in mind, too, that they could still be taking advantage VDI-like efficiencies, like on the fly disk streaming, rapid provisioning, single image management, and all the things that we commonly associate with virtualization, just without the virtualization aspect.

We keep getting nowhere with licensing

Dedicated hardware or not, I'm hearing murmurs that there is something custom going on between OnLive and Microsoft. Nobody is coming out and saying as much, but between conversations with some folks with a bit of inside knowledge and with Microsoft, we can start to see where the fog is coming from, even though we can't see through it. Questions directed at Microsoft were answered by saying they can't talk about specific customer's licensing. We got an equal amount of nothingness when asking about how we would go about building a similar situation on our own, though, which is odd.

Could OnLive be an acquisition target for Microsoft?

I thought it was odd that we couldn't get an answer out of Microsoft when positing a similar scenario, especially since you'd think Microsoft would want to encourage more Windows 7 adoption. Perhaps they're too focused on Windows 8 tablets, but running Windows on an iPad can't be too much of a turnoff.

Then I remembered Brian's article from last year in which he relays a conversation between him and Benny Tritsch about Microsoft's unspoken plans to use RemoteFX to deliver desktops, applications, and--gasp!--Xbox games from Azure.

Of course, this is uncannily similar to what OnLive is currently doing, minus the protocol and, perhaps, the virtualization. Still, with desktops and games from the cloud, each using the same protocol, and delivered from regional data centers around the world, I see some overlap. I wonder if Microsoft is working something custom with OnLive, and if this is kind of an experiment to see if it will work before bringing the into the fold. It could also explain the fuzzy logic around the Microsoft licensing.

One last thing about an acquisition that might be interesting - OnLive CEO Steve Perlman has already sold something to Microsoft once: WebTV. Not that it ever took off, but it does show that the relationships could still exist and that, this time, each company's offering could be on the same trajectory.

Browsium releases Ion to help migrate to Windows 7, Microsoft no longer pissed off

Gabe Knuth — Wed, 01 Feb 2012 05:01:00 GMT

Browsium has been around for almost two years now with their UniBrows product that runs an Internet Explorer 6 engine inside IE 8 and 9. In that time, they've found a number of ways to get on Microsoft's bad side. With Browsium's latest release--Ion--it appears they're friendly again.

If you're not familiar with Browsium, Brian recorded a podcast with CEO Matt Heller back in November of 2011, so you should check that out. Tim Mangan also wrote a blog post about them, and Shawn Bass mentioned UniBrows in a session at BriForum last summer titled "Coping with IE6".

Announced yesterday, Ion is a new take on IE 6 compatibility in a Windows 7 world. Ion replaces UniBrows which, while popular, was at best not supported by Microsoft. At worst...well...let's just say you can't buy UniBrows anymore. That's ok, though, because Ion builds on UniBrows to the point where you don't actually have to be using the IE6 rendering engine at all (that's the part that makes Microsoft happy again).

The reason Microsoft wasn't in love with the Browsium of a few days ago is that what they were doing wasn't necessarily encouraging people to abandon Windows XP, because they still needed the bits from IE 6 to work. Some companies were comfortable ignoring all those "rules" and "EULAs" and went ahead with the solution that worked for them and enabled them to migrate to Windows 7. Others just stuck with Windows XP. Neither solution sits quite right with Microsoft, so they worked with Browsium to create a better product. One that enables companies to migrate to Windows 7 without perpetuating ten year old, insecure bits.

UniBrows was limited in what it could do, so even though it was useful for sites that had weird javascript, old Java versions, or funky layouts, it still couldn't work with all the ActiveX controls out there because those still had to be interpreted by the modern browser. Ion, on the other hand, removes the IE6 rendering engine entirely. Instead it leverages the compatibility mode engines built into IE8 and IE9, which amounts to IE7, IE8, and IE9, plus Quirks, plus Adaptive Quirks, which is Browsium's additional engine that is NOT based on IE6.

Like UniBrows, this is all transparent to the user. Ion uses profiles to dictate which sites get which treatment. Those profiles are configured by the admin, and allow you to use different combinations of settings in different conditions. Some sites, for instance, require a certain version of Java combined with a certain rendering engine, while others have funky JavaScript that breaks in newer browsers. You can also configure various other settings that apps may depend on, such as group policy settings, environment variables, registry settings, and file locations.

With these profiles, Ion allows you to set up rules to deal with those situations based on URL strings. In the case of JavaScript, they actually have a pretty cool way of modifying strings with an on-the-fly find & replace that can replace any bad string with a good one. This string replacement feature and the Adaptive Quirks engine are what allow Ion to replace UniBrows. I'll be interested to see what changes, if any, people experience in the field as the migrate away from UniBrows. The admin features look really, really cool, and my notes are littered with little asides that say "this is cool" or "cool feature."

I'd love to test this software to get a feel for it, but this isn't exactly the kind of thing you can stand up in a lab--you pretty much have to have a real environment to test in. So, if you've already gotten your hands on Ion, let us know your thoughts in the comments.

In the meantime, though, it appears that Browsium and Microsoft have worked out a product that will both help the people that depend on Windows XP for IE6 and allow them to migrate to Windows 7 in a supported, legal way. Nobody can be mad about that.

BriForum 2012 London Call for Papers is now open!

Gabe Knuth — Mon, 30 Jan 2012 15:38:00 GMT

Last year we had our first ever BriForum in London, and it was such a great experience that we're bringing it back for 2012 (23-24 May). Part of the BriForum experience is seeing sessions from your peers that cover real-world, technical issues going on today. At BriForum, there isn't any rah-rah marketing and showing off technology that's six months away. BriForum is about linking the community with the information and answers that they need right now.

The reason BriForum is so amazing is that every single non-sponsored session is put through our Call for Papers process, which we use to select the best sessions for each BriForum. Each year we have some familiar names and faces on the list of presenters, but we can also boast that between 10% and 15% of our speakers are presenting at their first-ever show!

Like years past, we're accepting sessions on just about any topic related to desktop or application virtualization. BriForum is a technical conference, so the more technical and real-world your session, the better.

Session submissions are due February 24, 2012, and you will be notified whether your session was accepted or rejected on March 13.

The general experience for speakers is the same as years past. So if your session is accepted:

We pay your travel to/from London
We pay for your hotel while in London
We pay for your conference admission

If you're session has multiple presenters, we can only pay for the primary speaker (the person who fills out the form). We can sometimes work out a deal, though, to get you both there. (For example, both speakers get admission to the conference, but they have to share a hotel room and pay their own travel.) If you have any questions, please email Lindsay Jeanloz (ljeanloz@techtarget.com).

From a technical standpoint, here's what you need to know:

No more 1024x768 - go with whatever resolutions fits on the projector. Keep it realistic, though, since we don't really know what the projectors will be.
We will record your session. While you will own the content as your intellectual property, part of the submission process is you giving us permission to publish the video, etc…
There is no video capture box anymore, so no need to worry about all that extra stuff.
You do, however, have to have the appropriate adapters to connect your laptop to the projector. We also have a few spare laptops in case you need them.
All sessions are 75 minutes long.

If you'd like to present at this year's BriForum in London (there will be a separate announcement for the Chicago show, which takes place in July), you can submit your session information via this web form:

http://www.surveygizmo.com/s3/785825/BriForum-2012-London-Call-for-Papers

There are a few important notes about submissions before you get started:

If you want to submit multiple sessions, that's great! Just make sure you submit them separately, otherwise they'll get lost in the mix.
You're welcome to submit sessions on whatever you want, but of course we're looking for the content that has the most broad appeal.
The session must be submitted by the presenter. If you're not in the weeds enough to submit your own session, we probably don't want your session. Anything not submitted by the listed presenter will be ignored.
If you work for a vendor, you can still submit a session. However, you cannot submit a session about your company's product. (Example, Kevin Goodman, who works for VMware, and I did a session last year about User Installed Apps. That was fine. We would not have accepted a session from him talking about the advantages of VMware View, however.) This has been an issue in the past, so if you're unsure about this policy, please read this.

Below is a list of example topics as listed in the Call for Papers section on BriForum.com:

Server-Based Computing

Microsoft Windows Terminal Services / Remote Desktop
Citrix XenApp and the Citrix Access Platform
Smaller SBC vendors, such as Quest Software, Ericom, 2X, Jetro Platforms, Propalms, HOBsoft, Leostream, etc.

Application Virtualisation & Streaming

Microsoft App-V
Citrix XenApp Streaming
Symantec Workspace Virtualisation (formerly Altiris SVS)
VMware ThinApp
Smaller vendors, such as InstallFree, Xenocode, etc.

Desktop Virtualisation

VDI / Server-hosted desktops
Blade PCs / Blade workstations
Citrix XenDesktop, VMware View, Microsoft VDI bundle, Quest, Symantec, etc.
Client-based VMs and client hypervisors

Installing and Managing Centralised Applications

Operations Management
Application Streaming and Virtualisation
Business Impacts created by Server-based Computing
Cloud Computing (as it related to desktops and apps)
Software as a Service (SaaS)
Licensing
Scalability and High-Availability
Security
Scripting and Programming

Plus, we've branched into the Consumerization of IT, so anything effecting that space is also considered for BriForum this year.

Happy submitting, and hopefully we'll see you at BriForum!

Breaking down OnLive Desktop: Why this is not the desktop virtualization solution you're looking for.

Gabe Knuth — Wed, 25 Jan 2012 05:14:00 GMT

I've been trying to get together with OnLive for well over a year now, ever since seeing their cloud-based gaming platform. I'm not much of a gamer, though. For me it was the protocol that they use that generated the most interest. For a while my requests fell on deaf ears, but when they announced their OnLive Desktop solution in early 2011, followed by a release during CES in 2012, they joined our desktop virtualization world.

Jack happened to make a contact at CES, and it wasn't more than a week or two later that we had a meeting with OnLive CEO Steve Perlman down at their offices in Palo Alto. We went in armed with the questions that almost everyone has:

What protocol is OnLive using?
What platform do the games run on?
How is the backend brokering done?
Is OnLive Desktop using Windows 7 or 2008 RDS?
How is Microsoft Licensing handled since there's no SPLA license for Windows 7?

Since the meeting, I've had a chance to use the product and form some opinions based on what we talked about, what I've seen, and some speculation. What follows is what I know about these questions, as well as my thoughts on the experience and the product in general. Like me, you will still have questions at the end. It's still worth reading…I promise.

What protocol is OnLive using?

Sometime in 2010-2011, I'd heard that OnLive was using HP RGS as their protocol of choice. I actually met with HP earlier in the week, and they confirmed that HP RGS was *not* the protocol behind OnLive. As it turns out, OnLive is leveraging a combination of both their own silicon, their own UDP-based protocol, and partnerships with all the major ISPs in the country.

At the silicon level, OnLive touts a custom board in each of their servers that works in conjunction with their "System" (that's my word for their ambiguous system behind the scenes. It's a recurring theme). The System encodes the video stream as a full screen video, then sends it out via a UDP protocol. What's interesting is that the OnLive client is aware of the capabilities of the client (screen size, bandwidth, local resources, etc…), and OnLive says that they actually have a different protocol for each device. The example they gave was that with Android they get some access to the hardware, so they can send it a stream that encoded better because they can use the GPU or DSP to offload some of the decoding, whereas with iOS devices they have to do everything in software via the CPU, so the video stream sent to iOS is optimized for iOS devices.

Protocol aside, there are still a few challenges in delivering a realtime, 3D gaming experience (or a desktop, for that matter) over the internet. Part of that is addressed by the UDP protocol. As we know, UDP isn't subjected to the same checks as TCP. UDP packets are simply blasted out at the client with no regard for whether or not the packets were received. TCP, on the other hand, will retransmit lost packets to the point where it holds up the flow. In a game, this would be bad. Just talk to VMware about PCoIP :) Still, using UDP means that some packets might not make it to the client side. When asked how that works, I was told that there is technology on the client side that puts something in place to make it look normal. I imagine this like the predictive pixel technology that's built into 120Hz + televisions, where even though the signal is coming in at 60Hz, the processor in the TV knows what the first and second frames look like and can fill in the gaps. In practice, it looks like it's just compression, though.

The last thing that brings all of this together is the relationships OnLive has with all the major ISPs in the country. If you've read their marketing collateral, you'll know that there are bandwidth and distance limitations to using OnLive for games. Those limitations are different (less restrictive) for desktops, but they still exist because the protocol depends on those relationships to provide a good experience. Essentially, OnLive's systems plug directly into all of the major ISPs in the country, which they call "peering." This reduces the number of hops required to get from OnLive's servers to you. It's not a QoS thing, and OnLive isn't getting any preferential treatment on each of the major networks, it's just that the path the information takes is as short as possible. Part of the session initiation process is to identify the shortest path to the servers, and then lock that in.

Before we move on, let's sum up the protocol discussion. Custom graphics boards that feed a System that encodes the video data and delivers it via device-specific protocols on a network with as few hops as possible. It's complex, but if you've played any games via OnLive, it works.

What platform do the games run on?

I asked this question as an aside to my licensing question, and it turns out the answer is fairly complex. Unfortunately the licensing answer wasn't as comprehensive as the platform answer, so we'll go with the good one first. The platform, in a word, "depends." I went in thinking "Windows Windows Windows," because that's the licensing and platform that I cared about. The thing is, they have a backend that's capable of running games from different kinds of systems. Some of the games they provide run on Windows, some on Linux. OnLive actually gets custom games from developers that run on their systems, which makes sense when you consider the number of PS3 and XBox games they have, and the fact that the PS3 runs a Linux derivative. It's not like there are XBox and PS3 consoles sitting in a data center (well, there might be, but I doubt it).

How is the back-end brokering done?

This is a much shorter section, because the answer is "it's all proprietary." OnLive has spent a lot of time assembling their backend, and they claim to not be borrowing bits and pieces from other companies. We went in wondering if they use some sort of packaged or open-source brokering solution, but everything they've done is on their own. There are data centers around the country and in Europe, so when someone signs in, they're pointed to the data center closest to them,. Frankly, with the varied systems they run to support the games, a home-grown brokering solution makes sense. They're leveraging this system for the Windows desktops, too.

Is OnLive Desktop using Windows 7 or 2008 RDS?

Believe it or not, the answer I got here was "It depends," which I thought was weird because I simply asked "What version of Windows am I getting if I connect to OnLive Desktop? Windows 7, or Windows 2008 R2 RDS?" At this point it was like the doors closed and the information flow slowed. In practice, it was pretty easy to find out. I signed up for an account, launched the desktop, and poked around. It looks like Windows 7, but that alone isn't enough since it's possible to make Windows Server 2008 R2 resemble Windows 7. Most of the features were locked down (in fact, for the free version of OnLive Desktop, there isn't even a web browser), but signs throughout the UI point to Windows 7. I thought I'd try some old tricks, like File > Open, right click on CMD, have your way with the machine, but that didn't work.

I finally found my answer when looking at the Help > About screen in Notepad. OnLive Desktop is using Windows 7 Enterprise!

How can this be? Read on...

How is Microsoft Licensing handled since there's no SPLA license for Windows 7?

This is the biggest question I've heard from the people I've talked to because as many of us know, there is no Service Provider License Agreement (SPLA) for Windows 7. A SPLA program would allow organizations to deploy Windows 7 to any user at any organization (think DaaS providers), and it's something that the industry has been waiting for for years. As of today, the only way to deploy the same copy of Windows to users in different organizations is with Remote Desktop Services on Windows Server 2008 R2 (and prior versions).

Ignoring the lack of a SPLA license program for Windows 7, other considerations have to be made for client device access. Currently, OnLive desktop only works on iPads, which for obvious reasons can't have SA entitlements. That means that a VDA license must be purchased for each instance of Windows. Of course, that's not the proper method since we're all accessing this as a cloud service (unless it's running on dedicated hardware, but we'll get to that in a minute), but even if we were all OnLive employees, we'd still have to have that license.

I posed these questions to Steve several times, and never got an answer besides (and I'm paraphrasing) "it depends" or "that's not the hard stuff--the hard stuff is in delivering this experience."

I was told that OnLive has licensing experts that have all this worked out, but I was never told how that could be. I explained the licensing issues, the SPLA, the VDA, the fact that there are dozens of DaaS providers that are trying to accomplish the same thing but can't due to Microsoft licensing restrictions. Each time, my question was deflected and focus shifted over to the device demos.

I even asked about it in the context of games. I assume there are some Windows games in the OnLive library, and each of those would have to have a specific license for remote access as well. It has nothing to do with the fact that it's a game - it has everything to do with the fact that users are accessing applications running on Windows. OnLive has been in the works since 2005, and available to the public since 2009, but this is the kind of thing we've been dealing with for 15 years.

So, there are a few things that could be going on. The most likely solution that would be in compliance is if OnLive was using dedicated hardware for each instance of Windows. This would be entirely legal, but to do so could be very inefficient. It still means that VDA licenses are required, though, which isn't necessarily a small cost. It could be that they may have to run dedicated hardware for each session (game or desktop) to support the gaming side of the operation, in which case they probably have some super-efficient, motherboards-on-a-rack (or blade PC) infrastructure behind the scenes that makes it efficient to pull this off. If that's the case, then perhaps all of the above is true. I'm not necessarily sure if VDA licenses would be valid in this scenario, but I'll call on the licensing nerds to help us out in the comments on that one.

(Update: It's been suggested that you don't need VDA if the remote machine is installed on a physical host, so this is looking more likely. Too bad we still have to guess)

When giving away access to Windows desktops for free (or even $9.99 as the Pro plan will do), it's not easy to make up for that cost, let alone the infrastructure required to support dedicated hardware for each installation of Windows. In the meeting we had, they mentioned that had predictable usage patterns where the games were used more heavily at night and the desktops during the day. If that's the case, perhaps they're leveraging that, and the video game solution is subsidizing the virtual desktop solution (or, in effect, leveraging unused resources during the day). You'd have to think that remoting video games would be subjected to the same license restrictions, so maybe this is something they've already addressed.

Another, albeit less likely, option for a legitimate solution would be if they had some sort of special agreement with Microsoft worked out. If that's the case, Microsoft, there is a large, opinionated group of people that would love to talk to you.

Anything other than these two scenarios is likely not in compliance with current Microsoft licensing.

On top of all of this OS licensing talk, there's also the issue that both the OnLive Desktop Free and OnLive Desktop Pro apps come with Microsoft Office installed. How, again, can this be done? This is less complex than the OS licensing issue, but still one to consider when looking at the overall solution and the licensing cost associated with it.

What's OnLive Desktop like?

If you watch the video that's on YouTube of Steve Perlman giving a demonstration at the NExTWORK conference last summer (link to video, presentation starts around 28:30), you'll see that OnLive is very proud of their product. From a video game standpoint, they should be. They've pulled off an impressive feat by enabling realtime 3D gaming over the internet. I think, though, that OnLive overvalues that technology and just assumes that it will work just fine for desktops, too. In the video I mentioned, you can see Steve comparing OnLive Desktop to Citrix, saying things like OnLive Desktop is "...kinda like what Citrix is doing for the cloud thing…except the remote desktop feels local. In fact, in a lot of ways it's better than local," and, after essentially shadowing a session, he says "...and again, for those of you who've used Citrix, you've never seen anything like this before." He goes on to show a demo of a Flash app from SpeedTest.net running in a session, showing what great speed the datacenter has.

Even if I hadn't seen this before (I've been watching the same basic demo for 13 or 14 years), I'd still be amazed at the fact that to get these features you need to have a minimum 1Mbps connection to the web, and that it would be best to have 2Mbps. Frankly, if Citrix, VMware, or Microsoft's protocols had that much pipe all the time, their stuff would rock, too.

So, I tried OnLive Desktop on my iPad, and short of shooting a video demo of it, you can see where the technology they have that works fine for games falls a bit short when it comes to Windows desktops. The thing is, the compression algorithms used on games (which only go as high as 720p in OnLive's system) can be pretty liberal because there is motion all the time, and motion obscures what's happening on the screen anyway. Compressing that is probably imperceptible, or at the very least usable. With static windows that use very precise, fine lines, Windows desktops expose some of the compression techniques used by OnLive.

I don't know if it works this way on purpose or not, but when the low res, highly compressed image appears on the screen, you can watch it do what VMware calls "build to lossless." There are a few screenshots below. Keep in mind that these are just snapshots, and within a few seconds the screen changed from this to something useable. Also note that this did not happen all the time, but often enough over my 20Mbps home internet connection that it's worth mentioning.

Loading Excel

Excel loaded, spent a second or two like this, then turned "normal"

Another thing worth noting is that there is absolutely nothing in the local client in terms of user tools. Every other remote desktop client that we'd use on a regular basis has a client-side keyboard, scroll, right-click, and other usability features. OnLive Desktop relies entirely on Windows 7's built-in touch capabilities, remoting your touch commands instead of translating them into mouse clicks. This is cool because you get some of the multitouch and Surface capabilities and handwriting recognition that we've been hoping for the past few years, but in the end the user experience isn't as nice. Still, they almost require a bluetooth keyboard to make it usable, since it otherwise relies on the built-in Windows 7 keyboard app for typing (which means that you can't type well unless the connection is rock solid). Kudos to OnLive for the try, though. I'd love to find a way to get the best of both worlds when it comes to controlling your remote session.

So why does everyone else think this is awesome?

I think we were all sold on the "holy shit" awesomeness of being able to play 3D games online, and, like OnLive, our thinking was that this just has to be the ultimate solution for desktops, too. The general consensus of people in the desktop virtualization industry, though, seems to be "meh." I think the people that are drooling over this technology are the consumers, the users, the gamers, iPad fanboys, and anyone else who hasn't really been exposed to this technology before. To them, this is amazing technology, like having a computer somewhere else with a really long keyboard and mouse cable (or whatever simile you used to use to sell the tech in 1999). Today, though, we want perfect performance on almost any connection, on any device, and that means remotely rendered keyboards and super-lossy compression along with high bandwidth requirements aren't going to cut it.

What about the other versions that OnLive announced?

OnLive Desktop Free is available to anyone with an email address and an iPad by visiting desktop.onlive.com. It's limited to only a few gigabytes of cloud storage (which is owned and operated by OnLive, although they're looking to get out of that business), and it includes some Microsoft Office applications, but no Internet Explorer. It's also subject to system availability, so maybe there's a licensing catch in there, too, like they're only allowed to have a certain number of machines available in this manner. OnLive Desktop Pro, at a cost of $9.99/mo, will give you 50GB of cloud storage, "cloud-accelerated web browsing" (which means that IE is using their datacenter's connection to the internet, not anything as complex as Amazon Silk), and the ability to add PC applications when it comes out "soon." For both editions, non-iPad versions are also coming that will work on other devices, including OnLive set-top boxes.

Beyond those versions, OnLive also has plans for an enterprise version of the solution that will essentially give organizations access to the bare metal, onto which they can install anything they want. In this situation, it's all about using OnLive's network and compute power, and the customer assumes the software acquisition and licensing costs. This solution will be interesting to follow, since there are so many other companies out there that have DaaS solutions and hybrid cloud offerings. OnLive's key attractions at this point are their network peers and their protocol, and if people can get better performance in on more devices with more management capabilities elsewhere, OnLive Desktop Enterprise may never make it out of the gates.

Wrap Up

My tone is less than excited about this, as I'm sure you can tell. I love the product for the gaming aspect, but I think that OnLive needs to be more candid when talking about their licensing and protocol. I think it's amazing that they're using full-on Windows 7 but can't answer the question about how it's licensed. Ignoring licensing, though, I can still say that my brief experience with OnLive (compared to how other clients and protocols perform in the same scenario) leads me to believe that this is not enterprise-class desktop virtualization, and that the issues that affect online gaming are not the same ones that we have to deal with day-to-day to deliver applications to users. OnLive believes that if you can do video games, you've already solved the hardest problem in remote access to applications, so desktops are no problem. It sounded good initially, but I'm not sure I can agree with that anymore.

60+ BriForum 2011 videos now available to the public!

Gabe Knuth — Thu, 12 Jan 2012 05:01:00 GMT

It's that time of year again, the time when we start prepping for BriForum in 2012, and the time when we turn loose the videos from the past year's show. As of today (well, Tuesday if you were paying really close attention), all the session videos from BriForum 2011 have been posted in the Videos section of BrianMadden.com!

For us, BriForum is more than just a conference. It's a place where the entire community can get together, with little differentiation between attendees and speakers. Everyone is approachable, and each year we have new speakers that were attendees the year before. So, just as the community has given so much to BriForum, we like to give BriForum back to the community.

The videos available are from both BriForum London and BriForum Chicago. The London videos suffered from some snafus revolving around tape length (who uses 60 minute tapes for 75 minute sessions?) and video capture problems (who brings video cables that have broken red channels?), so there are only a few of them, but we had our best success ever with the Chicago videos, getting all but one, I believe. In total, there are 67 videos, including sponsored sessions (some really good ones, too). You can see all the videos from BriForum 2011 here.

As a technical note, the player used by these videos is our own custom Flash video player that we use to play two separate streams side by side. You can jump around in session using the scrub bar, swap the videos between the small and large window, and go full screen. You can't however, view this on an iPad or anything that doesn't support Flash. There is also no offline capabilities due to the backend complexity. However, if you want that capability, please let me know. The more requests we get for that, the more ammunition we have to revisit our video recording process and try to make it happen.

Happy watching, and hopefully we'll see you at a BriForum in 2012! Stay tuned for more information and for the Call for Papers process.

Who cares if you can't run "normal" Windows apps on an ARM-based Windows 8 tablet? It's all about future, not the present.

Gabe Knuth — Wed, 11 Jan 2012 05:01:00 GMT

At the Consumer Electronics Show this week, there’s been a lot of buzz created by Microsoft. Most of the activity surrounds consumer technology like Kinect, phones, and tablets, though, and not Windows 8 like I’d hoped. I have seen stories of a handful of demonstrations of the ARM version of Windows 8 running on tablets, and that brings to mind something important that is often overlooked: the apps you run in Windows on your Intel computer are not going to be compatible with the version of Windows that runs on ARM.

Many of you know this already, especially if you were around in the late 1990’s when Windows NT 4.0 came in four different flavors (bonus points for knowing which architectures were supported). ARM wasn’t one of them, but if you tried to install your run-of-the-mill Windows apps on an installation of Windows running on Alpha (there’s two…can you name the others? They're at the bottom of the article, but you probably already Wikipedia'd it), it wouldn’t work. Switching processor architecture means, at minimum, a recompile of the code, but more likely rewritten code.

Some people are upset, and I can understand why. They’re the people that don’t want an iPad or Android tablet because they can’t manage them the same way as a Windows box. They want to see a Windows box so that they use the same apps as on their desktop, and manage the OS the same way. Windows running on ARM sounds so promising on the surface, only to take it all away on a technicality. Two, actually.

The first technicality, of course, is the processor architecture, but the second is that the ARM version of Windows 8 reportedly won’t come with the traditional Windows desktop. While demos are still appearing with the so-called “Desktop App” accessible via the new Metro UI, other reports note that it will eventually be removed (Mary Jo Foley has been covering this all in depth).

Looking at the present day applications and management principles, you can see why they’d be upset. But if you look into the future, as Microsoft certainly has, it’s not hard to see where things are headed. People want the same apps and settings on all of their devices, no matter what or where they are. Installing things locally on the device is the most inefficient way to make that happen, even with sync. Microsoft is wisely shifting focus to the cloud, and the fact that they’re doing it with consumer devices first is no accident – consumers will be adjust and adopt a cloud mentality before big business will.

The Metro UI, which has shaped up to be more than just a UI, is also their approach to moving to the next generation of apps and services. While the kernel may be the same in Windows 8, it appears that the Win 32 interface known as the desktop and Metro UI are independent of each other.

It seems confusing, and it will until Microsoft finally gives us all the information we want, but piecing together what we know about the future of applications with what we know about Windows 8, we can see that the path they’re going down makes sense. If you have to write new apps to support the cloud, why not write them for an architecture that’s optimized for mobile devices? ARM processors are found in a whopping 90% of mobile devices, from phones to tablets. They’re preferred for their high efficiency, low power consumption, and flexible production. ARM doesn’t actually make processors – they license the architecture to chip makers who design their own chips around the ARM technology.

ARM processors are expected to enter the PC market, too. In 2011, IDC projected that ARM processors will be in 13% of PCs in the next three years. These PCs aren’t going to be running Windows and Office like they do today. They’ll be cloud-oriented desktops with some local apps, like a web browser or client application for a cloud service. They’ll be tablets with external monitors, keyboards, and mice. The reason for this is that the world is moving in this direction, and when it gets there, you won’t need that overpowered muscle car under your desk—just the sleek, efficient hybrid.

So, while I can understand the furor over apps not being compatible with the ARM version of Windows 8, I think those that are upset need to take this opportunity to look at what the future of applications and operating systems will be. There’s a very real possibility that ARM could become the default architecture of the future, and that Windows 9 or 10 will just support x86/64 processors as a necessary evil. Something major could happen, of course. Intel is talking about their mobile chips, but unless they support the same apps (ARM-based apps) as everyone else, it’s a lost cause…just like trying to run Windows apps on a tablet.

Oh, and the four architectures supported by Windows NT? Alpha (DEC), x86, MIPS, and PowerPC.

If Apple allows iOS apps on Apple TV, will that be the ultimate thin client for only $99?

Gabe Knuth — Wed, 14 Dec 2011 05:01:00 GMT

We've all heard of the Apple TV. Some love it, some hate it, and the reasons for that are pretty well documented. The thing is, most of those opinions all pertain to its ability as a home theater device and not much else. Still, it's hard to look at the Apple TV without wondering what else it could do. With some modifications, I think Apple TV could be transformed into a powerful, reliable, and inexpensive thin client. It sounds like the ramblings of a fanboy, but read on, then let me know what you think in the comments.

First, let's look at the device specs. The latest iteration, released in 2010, runs the A4 processor that was included with the first iPad and iPhone 4. While not as powerful as the dual core A5 or the rumored A6, it's no slouch. It also runs iOS, albeit a locked down variety. It's been jailbroken, though, and applications do exist that can run on it outside of Apples oversight. It has HDMI video out (720p only, though, so maybe not capable of displaying a useful, monitor-sized window) as well as bluetooth hardware. Bluetooth is not enabled by default, but it wouldn't take much effort for Apple to turn that on (it's been done by jailbreaking). It also has a real, wired ethernet port, so it doesn't have to rely on WiFi.

So, we can connect monitors already. If Apple opens up the hardware a bit more to allow connecting keyboards, mice, and wireless headsets, while opening up the software so that we can run regular iOS apps on the device, we could take advantage of all of the clients and applications out there to truly have a $99 thin client. Let's take a look at some of the capabilities we'd have:

Clients for just about any remote desktop

VMware, Citrix, Quest, Ericom, Virtual Bridges, Wyse -- everyone has a client that runs on the iPad, and since the hardware is similar, it stands to reason they'd run just fine on AppleTV, and with each solution's native protocol. The only thing I can think of that isn't supported right now is RemoteFX, but who's using that in a thin client environment. There are some, but not many.

Centralized Management

Mobile device management is H-U-G-E right now, and the solutions that people use today, Good, Matrix42, System Center, OK Labs, and so on (boy is that an incomplete list!) would easily work with Apple TV. They're fundamentally about managing the OS on the device, so it's not too much of a stretch to use them in this case. The device/OS can be managed, as well as the local applications that can be accessed by the users.

Corporate iOS apps

Apple actually has the ability to work with you to deliver your company's home-brewed iOS apps just to your users for local execution. If you don't want to write native apps, you can still deploy them via the browser since it could have the HTML5-compliant Safari browser built-in. Of course, companies are building and launching iOS-compatible versions of their clients all the time, so you could just deploy those, too.

You might mention that there are inexpensive solutions in the works for Citrix and RemoteFX. Citrix just announced the HDX System on Chip at Synergy in Barcelona, and Microsoft has been talking about the RemoteFX thin client for as long as they've been talking about RemoteFX. Neither is out, but prototype hardware does exist, at least. The thing is, those products, and anything else like them, are relegated to a single platform. Using Apple TV, much like our speculation around Android thin-clients, would allow orgs to work with any platform. In the case of Apple TV, though, the hardware is there, known, already being developed for, and supported by almost everyone.

It's not all about the price, and $99 might be too low for this, given that for Apple to expose the OS and the device would probably require more support and work on their side. Flexibility contributes to the overall value here, too. All of those clients, capabilities, and protocols could be supported on one device that already exists.

ControlUp - The free management tool I wish I had when I was a full-time admin

Gabe Knuth — Thu, 08 Dec 2011 05:01:00 GMT

Have you heard of ControlUp yet? Smart-X, its creator, has received a bit of press lately for it due to it's simple, unique approach to monitoring and management. I had a chance to meet with Yoni Avital and Asaf Ganot when I was at Citrix Synergy in Barcelona, and while Control-Up isn't going to be everything to everyone, it is a free monitoring tool with some snazzy features.

Like a lot of people, I'm likely to glaze over when it comes to monitoring apps. VMware's Kevin Goodman coined the term "YAM" a few years ago, which is a TLA (Three Letter Acronym) for "Yet Another Monitor". There are probably a dozen, if not more, monitoring solutions in the desktop virt space, and while some are very comprehensive (and complex), others are just dashboards that find new ways to display PerfMon counters. There are exceptionally powerful ones that use their own technology, but in many every-day situations, those solutions can be overkill.

ControlUp takes a different approach. As it is today, ControlUp runs entirely in memory. On the server side, the management console collects data from lightweight agents that have been pushed out to the clients using RPC. There's little impact on the client side, but since all of this is not cached or otherwise stored for later, the memory impact on the server side can be substantial. Nothing a typical box can't handle, mind you, but you're probably not going to roll this out in production on a Windows 7 VM with 1GB of memory.

What I like about ControlUp is the real time aggregate view of the organization's computers. It works on physical and virtual desktops, including both VDI and TS-based. From the management console, you can get sortable views of every single process running on monitored machines, sessions, accounts, and executables with process-specific and aggregate CPU and Memory information. From there, you can also dig into some of the more subtle features of ControlUp, all of which can be done on single or multiple machines at once, in real time:

Group policy refresh
Filterable event log aggregator
Session screen shots (Check out the comments on this blog post from Citrix)
Manage Windows services
Copy files/folders
Registry edits
Shutdown/reboot
Remote Desktop Connection console to access multiple computers via RDP from one interface

Click for a larger view

While many of these things can be done with scripts or existing tools, ControlUp gives you the ability to do it from a single console, for free, and without the complex backend to worry about installing or maintaining. If you want oodles of metrics saved off to a massive database to correlate events with historical data, this isn't for you. But, if you've got a bunch of homemade or disparate solutions to accomplish what ControlUp enables you to do, check it out. That's what captured my interest. During the demo, I found myself thinking back to how much I could've used this (free) app back in my days as a full-time admin.

Things to note:

I'm not sure how well some of the features scale, especially when dealing with less-than-optimal networks. Things like file copies could be problematic, I think, but I could be wrong.
Control-Up does require you to have an account on the Smart-X website to verify your license (even though it's free) and collect usage stats. It seems legit, but unnecessary.
There is no offline way of using this as an ongoing monitoring solution. When you shut down the management console, all the info goes away.
Smart-X is working on an enterprise version of Control-Up, so if you have ideas for features you think it should have (like offline use, database storage, etc…), let them know. Many features, like being able to turn off the desktop screenshot feature, have come from user requests.

For more info, check out the ControlUp website or watch some demos on their YouTube Channel. If it's not quite where you'd like it to be, get in touch with them and let them know how they can enhance the product. They seem more than willing to listen.

Citrix shares a bit more VDI in a Box vision. Bottom line: 2012 will be a better year with a consistent message

Gabe Knuth — Wed, 30 Nov 2011 05:01:00 GMT

I had a phone call with Kumar Goswami, formerly of Kaviza and now VP of Product & Strategy at Citrix about the article I wrote a few weeks ago on Citrix's VDI in a Box message. It turns out that, while some of the field messaging that we've been hearing is off-base, the message that Kumar is delivering within Citrix is pretty close to in line with what I wrote about.

Kumar agreed that the use case should determine the best solution for the organization, regardless of the size of the organization. The only place that company size comes into play is when spreading leads around to partners. In that case, they're using company size to determine which product should be used to break the ice with a customer, not as a technical restriction.

That could still be a problem if existing preconceived (or ill-conceived) notions are believed, so to deal with that, Citrix is starting to ramp up an education program for marketing and sales (which also flows down to partners) on just how to decide on the proper product for a use case. Part of that effort will be to help identify situations where VDI-in-a-Box works in large organizations with an "SMB use case" (I'm still not a fan of calling in that, but I guess you have to call it something), and vice-versa.

Kumar also explained 2011 has been spent onboarding the Kaviza team and getting the product Citrix-ified (my word, not his :). Because of this, marketing and sales haven't gotten the attention and education they need. That process begins soon, and by early 2012, they expect the "use case" message to be consistent across the board. That should deal with all the different numbers being thrown around regarding how many users VDI-in-a-Box is meant to handle.

With regards to an upgrade path to XenDestop, there a number of technical reasons that an actual software upgrade can't happen, but Citrix is hammering out the details of a license-only upgrade that will essentially allow people who have invested in VDI-in-a-Box to put that investment towards XenDesktop licenses in the future. While there is no formal package yet, Citrix is dealing with the issue on a case by case basis. So, if you want to take advantage of the upgrade, contact your rep.

A few other notes that tie up some loose ends from other conversations, not necessarily from the article:

VDI-in-a-Box will be in the CSP (Citrix Service Provider) program next year, which is a program that covers licensing for hosted environments.
There are no known caps on how many users/sessions/desktops VDI-in-a-Box can support. The plan is to let the market figure that out.
They realize that there is some overlap between the two products, but the hope is that education can narrow the gray area

Hopefully 2012 bears this out and I can finally find a new soapbox. And, CSP's - I'd love to hear how how both XenDesktop and VDI-in-a-Box work for you once everything becomes official. As you may know, Brian is becoming a huge fan of outsourcing VDI desktops for up to thousands of users, and I'd like know what you can deliver easily and what your main challenges are when moving desktops outside and organization. We'll save that for later in 2012, though.

Citrix needs to change its strategy for VDI in a Box

Gabe Knuth — Wed, 16 Nov 2011 05:06:00 GMT

I'll start at the end - Regardless of what Citrix tells you, don't let the size of your organization dictate whether or not you use VDI in a Box instead of XenDesktop. The real reason for using one or the other should have absolutely nothing to do with the size of the organization or the number of VDI seats. Like choosing any VDI platform, it should be based entirely on use case.

Many conversations I have about VDI in a Box include a discussion about how it's slated for XX number of users. I had conversations with just about anyone who would listen at Citrix Synergy in Barcelona a few weeks ago, and I learned that while Citrix partners are touting VDI in a Box as a sub-500 user solution (or sub-1000, depending who you talk to), in some places in Europe, it's being sold as a sub-50 user solution!

To me, this means that, not only is Citrix conveying the wrong message about how to use VDI in a Box, they're also allowing partners (and possibly themselves) to use arbitrary numbers. That leads to internal competition between two completely separate VDI products. Worse, it means that sales reps and partners are left pushing the wrong product because of higher commission or margin (not that that's anything new), which isn't best for the customer. This kind of confusion is on par with that of Microsoft licensing.

For a few reasons, this all gets me pretty fired up. First, I've always had a soft spot for Kaviza. They've been an underdog with a really tight, simple solution. To see them get acquired just to compete against another product in their new company is not exactly what I hoped for. Second, while I was happy they were acquired by Citrix, it was with the hope that the simplicity of VDI in a Box would rub off a bit on XenDesktop, and that the two would either blend into a single, harmonious product or, at the very least, VDI in a Box would serve as an entry-level VDI product with a clear, easy upgrade path to XenDesktop.

So how can Citrix and its partners get it right?

In the short-term, Citrix can stop focusing on the number of users for the project. Yes, it's a factor, but so many other things come into play before number of users. Things like storage requirements, persistent vs. non-persistent VMs, and the need for FlexCast (which amounts to XenDesktop, XenApp, XenClient, and streamed apps) all come into play just as much.

I mentioned this to a few people, and I was told that this is mainly used as a guideline for the company size to target (SMB for VDI in a Box, enterprise for XenDesktop). To me, this is even more useless, because now we're talking about targeting only sub-1000 person companies for VDI in a Box, and only 1000+ companies for XenDesktop. That means that a company might need XenDesktop or VDI in a Box, but never learn of it because the partner they're talking with leaves it out of the conversation based solely on company size.

Note: I do realize that this is how Citrix can say that the products don't compete against each other. If you say VDI in a Box is only for sub-1000 person companies and XenDesktop is for 1000+ person companies, then there's no competition because they're meant for different areas. As you can imagine, I don't think that's true :)

If you have an environment that doesn't require, say XenClient or XenApp and local storage with non-persistent VM's is ok, then you should absolutely be leaning towards VDI in a Box. Only then should you consider the number of users to see if your specific use case will scale properly. There are implementations of VDI in a Box that go well above 1000 users, so the precedent is already there that it will scale to accommodate the size of the vast majority of VDI deployments in the world.

The opposite is also true. Say you only need fifty users (regardless of company size), but your use case requires XenApp and XenClient, along with built-in App-V support and has high storage requirements. XenDesktop is probably going to be the product for you, even though it falls well below the 1000 (or 500) seat threshold.

So, it all depends on each specific use case, and that, to me, is the responsibility of the partner. Instead of selling the same solution over and over again (which is easy), they need to roll up their sleeves and actually discover the best solution for each customer. The problem is that the information their getting from Citrix is shortsighted at best, and downright misleading at worst.

How can this get worked on in the long term, then? I think Citrix has to get away from having two separate products that, let's face it, compete with each other. The "number of users" limitation is there to draw an arbitrary line between the two so that they can say they don't compete with each other, but that's pretty transparent (or at least it should be now). There are many good things about XenDesktop, but it is a very complex solution (a fact that VMware is very willing to reinforce). VDI in a Box, on the other hand, is perhaps too simple, with scalability concerns about it's grid architecture.

Citrix would do well by trying to adapt the grid architecture to be more scalable (or at least prove its scalability) and incorporate that, along with the ease of standing up new servers, into XenDesktop. XenDesktop can contribute by adding in the FlexCast features and Provisioning Server, which will help the image creation and maintenance processes where VDI in a Box catches some criticism.

There's probably a dozen ways to slice up the two products and repackage them, and it's not an enviable task. I do think, though, that it's a necessary one. Having two products that compete against each other in a niche market cannot be a good thing.

In the meantime, though, I'll bring it home. Forget anything you've heard about what size organizations should be using XenDesktop or VDI in a Box. Bring them both in, and put them through your specific use case. Compare your storage, VM state, application delivery, user base, and scalability requirements to each solution and let that determine which product you use.

For more discussion on this, check out the podcast of the Brian & Gabe Live show that we recorded yesterday at about the 12 minute mark. Desktone's Peter McKay was our guest, and he and Brian shared their thoughts on this as well. They bring up how small business don't want anything to do with owning and running their own VDI environment, let alone deciding on a use case, among other things.

Can Microsoft SA be a cheaper option to VDA for thin clients? If the stars align, maybe...

Gabe Knuth — Tue, 08 Nov 2011 05:21:00 GMT

Somehow, both Brian and I were on a licensing kick this week, and I was surprised to see his article yesterday about Windows Server 2008 as a way of skirting the SA/VDA requirement at the same time I was knee-deep in Microsoft licensing fun. For me, it's not about skirting SA/VDA, but more about trying to find the most cost effective program when using non-Windows clients.

Ask most people about what you need when using thin clients, and you'll hear "Thin client…yeah, that's VDA." I was talking to Jack last week after a conversation he had with ClearCube (yeah, they're still around…just a bit quiet outside of high security environments), and they had an interesting offering with their thin clients where they'd either:

Sell you a thin client that you'd have to buy VDA for
Sell you a thin client with a real Windows license that has SA

The first option is the no-brain answer. Since the device doesn't have Windows and/or wasn't purchased through the normal channels, you wind up having to buy VDA for each device at $99/yr.

The second option is interesting to me. Essentially, you're buying a Windows license for which you'll buy SA (at 29% of the cost of the license every two years), and throwing that license in a drawer somewhere since you're not actually going to install Windows on the thin client. It sounds crazy when you consider the acquisition costs ($99 for VDA, or ~$200 for a Windows license + SA), but it might be the most cost effective way to get your devices access to Windows VDI images. While the initial cost would be higher, the year over year cost would be lower. Please keep in mind that the pricing here may not represent what you actually get from Microsoft, and that anything I've come up with is essentially a SWAG - Scientific Wild-Ass Guess. I tried :)

What ClearCube is doing isn't all that groundbreaking, but it's a unique take on a problem that plagues everyone in this space - making sense of Microsoft licensing. I decided to take a look from a generic perspective, looking to use any device that does not currently have SA as a VDI client. I made two charts, one with retail pricing for Windows 7 Professional, gleaned from Amazon.com, and another with estimated pricing for Windows 7 Enterprise. I'm leaving the hardware costs out of the picture, since those are what they are. This is only to compare two ways you can get SA privileges to run/access Windows virtual machines.

Chart 1: VDA vs SA, Retail Pricing, Windows 7 Professional

This assumes a Windows 7 Pro price of USD$249, with 29% of that due every two years for SA. This is a bit of an oddball solution, though. From what I understand, you can buy SA for a retail copy of Windows, but each copy has to have its SA subscription activated and maintained separately, as opposed to part of an agreement. So, while there are visible cost savings after four years, there is a lot of administrative overhead maintaining the subscriptions.

Chart 2: VDA vs SA, Agreement Pricing, Windows 7 Enterprise

This assumes a Windows 7 Enteprise price of USD$150, with 29%, or USD$43.50, of that due every two years for SA. Of course, to get Enterprise, you need to have an Enterprise Agreement with Microsoft. Maintaining the agreement is already being done, though, and the cost per license is significantly less expensive. If you have an EA, this is for sure the way to go.

There's one other option that perhaps thin client manufacturers could use, and what I suspect ClearCube uses: The COA license. The COA, or Certificate of Authenticity, is relegated for system builders. Now, you can go to Amazon and buy a COA license of Windows 7 Professional for $139.99 (only slightly less than my estimated Win 7 Enterprise cost), but you'd only be allowed to use it if you were building a machine from scratch (and, I believe, for resale). Still, this is something that thin client manufacturers could offer to help their customers avoid purchasing VDA. And, since the OS would never actually get installed, it could be whatever the lowest (cheapest) possible edition of Windows is that allows SA.

While you can't technically get COA licenses for all of your VDI clients, thin client manufacturers could, I think, help out with that by bundling them with their devices. I'm reserving the right to be totally wrong on this one, though :). I believe that there is some sort of restriction that only allows you to buy a Windows license with SA for a device that can actually run Windows (even if it doesn't run it, it has to be able to run it), too, which means you probably can't get away with buying a COA (or any other Windows license) for your Xenith thin client.

So what does this all amount to? It's pretty obvious that buying SA is significantly cheaper than buying VDA after a relatively small amount of time. Even against retail pricing, the cost savings start to appear in Year 4. Against OEM pricing, though, the savings are almost immediate, and tangible in Year 2. The savings of having SA instead of VDA after four years in the OEM scenario nets a savings of $159 per user. Stretch that out to six years, and the savings is $314 per user ($594 for VDA, and $280 for SA).

The question is, how doable is this? A lot has to happen for this to pay off, it appears. To get the huge savings, you need to be operating under an enterprise agreement, have clients that can run Windows (even if they don't actually run it), and be able to acquire the Windows licenses for those devices. If you can satisfy all of those, you can save a lot of money on licensing. This is also helpful for planning out your thin client deployment, because purchasing more powerful devices could end up costing less money in the long run if they can run Windows and save you the cost of VDA.

Let me know what you think. I could have this twisted in my head, but I've tried to go over a few times. I certainly wouldn't be the first to get confused by Microsoft licensing...

Citrix Synergy Barcelona Keynote News and Notes

Gabe Knuth — Wed, 26 Oct 2011 11:00:00 GMT

Good Morning! While you may just be getting to the office, I've been awake for a few hours, spoke in a session, and attended the Citrix Synergy Barcelona Keynote. This year, there we announcements aplenty. So many, in fact, that not all of them made it into the keynote (and one that did make it in, even though we could've lived without it).

Right off the bat this morning, Citrix announced the acquisition of AppDNA (you can read my impression of the move here). The short version of my take on this is that it's good, especially in light of the announcement of a partnership with Lakeside Software and the new Virtual Desktop Assessment tool. This means that Citrix has the tools to help assess your applications and your hardware/resources as you move on from traditional desktops.
After that, Mark T. gave a rundown of XenDesktop usage in the world, proclaiming it the number one solution on the market. One of the interesting notes is that he said there are 75 customers with 10,000 or more licenses! That's a staggering number, and it seems unrealistic until you consider that the vast majority of them probably took advantage of the XenApp to XenDesktop trade-up program. That means that most of them are still simply running the XenApp portion of XenDesktop. Is that number true? Yes. A little misleading? Also yes.

Citrix now has a firm footing on the Post-PC era message, and it didn't take long for Cloud to come up during the keynote. Consumerization also came up fairly often, with the message being that we cannot fight consumerization, so organizations need to embrace it. Naturally, Citrix is there to help :)
They continued the same message from Synergy in the US with regards to Total Value of Ownership. The change this time is with me. Back in May, I thought it was too subjective of a term, but the reality is that there is much value in desktop virtualization, regardless of whether or not it saves money. Brian's favorite analogy here is something along the lines of "you don't drive the cheapest car you can buy, do you? You pay more for more features." I'm 100% on board with that now, and with Citrix's message on it. Their cost models leave something to be desired, but the trend is still there.

(It's tough to explain, but it's not a very detailed chart, is it?)

ShareFile was up next, and there was a brief demo of the Outlook plugin. This plugin integrates their cloud storage solution into Outlook for attachment storage and such, and is one of the more popular features of the ShareFile offering. What's odd to me is that they showed cloud storage integration in the Citrix Receiver later on, but they didn't attribute it to ShareFile, even though it clearly is. Maybe it was just an oversight, but I think it would've been beneficial to show the quick turnaround of ShareFile features into Citrix products. The version of Receiver with the cloud storage/ShareFile integration will be available in Q4 in Tech Preview form.
GoToMeeting received another update with the "Workspaces" feature. Workspaces allows meeting participants to share a document in the meeting interface, but only in an upload/download kind of way. Collaboration is limited to making side notes a la Microsoft Word, while one person still needs to do the editing. I like the direction this is going, but this version falls a bit short. I want to see something that involves ShareFile sync and Google Docs-like real-time editing.
So long Web Interface, it's been fun. Citrix Cloud Gateway (specifically the Receiver Storefront portion) has replaced WI as the backend source of Citrix-delivered applications. Web Interface will still be around, but will not be revved beyond the current version 5.4. Same goes for CSG. No word yet on whether or not CSG will work with Citrix Cloud Gateway or if you'll be forced to purchase a CAG or other solution.
Also on the Cloud Gateway front, there will be two versions: Cloud Gateway Express and Cloud Gateway Enterprise. Cloud Gateway Express is simply the next evolution of Web Interface, with mostly the same feature set. It can be upgraded to Cloud Gateway Enterprise, which adds the ability to plug in more functionality, like SaaS app support. Citrix and VMware are truly in a dogfight here.
There's been a large focus on HDX, and I even tweeted something to the effect of: Citrix can't be accused of ignoring HDX. Every time we have attend these shows, there are more improvements. Today, the main focus was on a low-cost thin client that was developed through a partnership with Texas Instruments and nComputing. This thin client will use System on Chip (SoC) technology to offload processing of HDX features to hardware, and is aimed to be a low cost (assuming ~$100) solution. Citrix uses this to justify the decreasing device cost in the cost model chart above.
Last, we have the announcement that I wish we hadn't seen. We were teased with the "One more thing," and talks of social this and social that. In hindsight, we should've seen it coming, but out of nowhere we were shown a Citrix Receiver for Facebook. You could sense the collective eye roll in the audience, or at least in the blogger area. Maybe I'm getting old, but I feel like that's a blending of two worlds that just shouldn't exist. I guess we'll see what the overall reaction is to it before counting it out, but the general sentiment around me was "Are you f-ing kidding me?"

I'm here for the next few days with a video camera, and there is plenty to talk about and see. If I missed something in this rundown, let me know and I'll look into it for you. Otherwise, stay tuned for more from the expo hall floor here at Synergy 2011 Barcelona.

Citrix acquires AppDNA - Here's my initial take

Gabe Knuth — Wed, 26 Oct 2011 09:01:00 GMT

Citrix has announced the acquisition of AppDNA, in what has been one of the worst kept secrets of synergy this week. The acquisition itself is interesting, because AppDNA's AppTitude technology was considered to have a limited lifespan. Since their core product is related to migrations from XP to 7, it stands to reason that AppDNA as it is today would have little to no purpose after 2014 in a post-XP world (assuming they didn't change their product offering).

So why would Citrix want to buy a company for two years? There is still a large amount of companies out there deploying Windows 7. There are even some organizations (foolishly) waiting for Windows 8. Leading up to the announcement in the keynote, Citrix CEO Mark Templeton talked about helping with transitions and reducing the pain of change and adaptation. AppDNA's solution, combined with the recently announced virtual desktop assessment tool released in a partnership with Lakeside Software, now gives customers a definitive path to move away from XP. They now allow you to plan all aspects of your XP to Windows 7 migration in a predictable, reliable way. It establishes a trend started by Quest Software with their partnership with LWL and recent acquisition of ChangeBase.

So what about VMware?

It looks like VMware is now left in the dark when it comes to this trend, although they could choose to jump on this train with a similar product like Flexera. Odds are, though, that VMware would rather focus on moving users from XP to something more SaaS oriented. When both Citrix and Quest are betting on Windows past 2014, though, it would be foolish of VMware to ignore the idea. It's not like Citrix and Quest made these acquisitions without talking to customers and assessing their willingness to move away from Windows entirely.

Stay tuned for more and for an video with AppDNA that I'll shoot later in the day.

Waiting for 8: Why waiting for Windows 8 doesn't make sense

Gabe Knuth — Mon, 03 Oct 2011 04:00:00 GMT

Last week I read an article on Ars that suggested that companies might be considering keeping Windows XP until Windows 8 comes out. For the record, the article is dismissive of the idea, and I have to say, I am too. The main reason for me is that Windows 8 is still a ways off. Even though some sites are saying it's entered the "home stretch," keep in mind the release date is scheduled in 2012, and given the fact that the only installable version available now is pre-beta, you can bet on later 2012 than earlier.

See the problem there? April 8, 2014 is the date that Windows XP will be discontinued forever. I'm sure people will use it as long as the activation servers respond, but the vast majority of the world will have needed to move on by then. If it's taken them this long to decide on Windows 7 (which is two years and one service pack old now), I hardly believe that 18 months is a long enough time to overhaul systems and commit to Windows 8, even if it isn't as fundamental an upgrade as XP-to-7 is.

But what if the reason organizations are holding off has little to do with Windows at all? It wouldn't surprise me one bit to learn that companies are sitting still in spite of a mature Windows 7 and a one-year wait for Windows 8 because they're waiting to see how the cloud landscape shapes up. We all know there's enough buzz! VMware has all but stricken Windows from their future plans, and there is an ever-increasing gap between the Window and Window-less groups.

That's not to say there aren't huge Windows proponents out there that will continue developing for it and doing great things for as long as they can. Let's face it, though...for that group to remain relevant, Microsoft has to hit a home run with Windows 8. If they pull a Star Trek movie-esque sequence where every other release sucks, Windows, Windows apps, and traditional PCs could suffer a serious blow.

So where does Microsoft need to get it right? All signs point to Microsoft expanding their embrace of the cloud, mobile devices, and desktop virtualization. Windows 8's Start Menu really does look like it will be nice on a tablet, but will it be accepted as a desktop computer interface? What about in the corporate world, where a major change requires retraining and lots of user complaints? The desktop is still there, although the default interface I've seen so far is the tile-based Start menu. Remember when Microsoft switched Office to ribbons? It was a wholesale replacement with no option to revert to the old way. How long did it take some companies to switch from Office 2003 because of the retraining and awkwardness involved? Imagine that awkwardness for accessing all of your applications? I'm holding out hope that Microsoft keeps the Windows 7-style start menu as an option, at least in the business versions of Windows 8.

Hyper-V components will be included, reportedly, which I think is a good thing if implemented in a way that gives third-party ISVs the ability to use the hell out of it. It's so promising that I can see organizations waiting for it just to see what happens, but if Windows 8 is still a year away, we have to be at least 18 months away from a really compelling product that takes advantage of it.

So should organizations "wait for 8?" No matter what you're waiting for, I think it's foolish to wait for something that doesn't even have a firm release date. Upgrade to 7, if for no other reason than because you simply have to do it. You're going to have to upgrade Windows at least one more time before evaluating a complete shift in the way you access applications, even with some of the "bridge technologies" that are coming out like VMware Horizon and Citrix Cloud Gateway. Plus, if you wait for Windows 8 and Microsoft misses, Win 8 SP1 could be too late to make it right. In that case, desktop virtualization or not, you'll be left with an emergency upgrade from XP, rather than a controlled rollout.