Simon Bramfitt

XenApp Deathwatch (updated)

A little while back I suggested that Citrix should stop selling products and start selling services, specifically an app delivery service that and a desktop delivery service combining the best elements of XenApp, XenDesktop, Provisioning server for Desktops as a single suite that could be used any way that the customer wanted to deliver either a desktop or an app, hosted or streamed whenever and however the customer wanted.

Well today it seems that Citrix have gone one better. Not content with giving XenApp admins access to XenDesktop to host desktop OS dependant applications, Citrix announced today that the VDI guys can now have piece of the application delivery action by integrating XenApp licenses into what is fast becoming Citrix’s new flagship product XenDesktop 4. Now with XenDesktop 4, one single product license allows you to stream or host a desktop or an application anyway you want. I just hope Citrix don’t forget where the idea first came from - I’ll be expecting a check in the post Mark. 

Aside from opening the door to total delivery flexibility, the other really big change is that for the first time ever Citrix has stepped back from the concurrent usage licensing model.  XenDesktop 4 licensing follows the Microsoft model of one license per user.  The reasoning behind this is not clear and is likely to create some angst amongst Citrix diehards, but old time WinFrame admins are well versed in adapting to change. It will be especially difficult for the channel to explain why XenDesktop 4 is priced per named user while XenApp and XenDesktop 3 are both still priced per concurrent user. 

There's still one piece of the picture missing today, but with a clear indication that XenDesktop 4 will integrate with XenClient just as soon as it is launched, Citrix will have a complete application and desktop delivery suite available mid 2010 just in time for its customers to have got to grips with the idea that desktop virtualization is about to play a big part in their future.

I can’t say that I’m 100% happy though; XenDesktop 4 is still only licensed in terms of pre-paid licenses. There’s no service based offering providing capacity on demand or monthly billing, so you still have to get your demand forecasting right and endure that you don’t over or worse under estimate the number of licenses you need. But there's plenty of time for that in the future.

Here's Tom Flink explaining the announcement to the sales channel, makes me want to run out and buy it right now :)

So is this change the beginning of the end for XenApp? I’m sure we will see XenApp 6 announced towards the end of this year, bringing with it a raft of new management features, (Update  - I have solid assurances from Citrix that XenApp is not going to go away and that XenApp will continue to be enhanced well beyond XenApp 6) but as far as XenApp' loosing it's crown as Citrix’s flagship product is concerned; I think the answer has to be that yes we are seeing a transition here. I’m not sure we will get to see the passing of the torch acted out on stage during next year’s Synergy keynote, but I can see a few moist eyes in the near future.

XenDesktop 4 will be available Nov. 16. Prices range from $75 for Standard Edition (XenDesktop only), through to $225 for Enterprise Edition (XenDesktop + XenApp) to $350 for Platinum Edition which includes full management and security capabilities. XenApp customers with Subscription Advantage will be able to trade up to XenDesktop 4 with discounts of up to 80 per cent. 

More later …

VM Hosted Apps - Isn't it time we had a little simplicity?

A little while ago I was asked to consult on a rather nasty system integration problem that was causing an XenApp migration project to come unstuck – an application that had been working without problem for over 5 years had lost the ability to support multiple concurrent users, and what ever the problem was it was specific to this one application.  Other apps on the same set of servers were fine and needless to say aside from the migration from MetaFrame to XenApp 'nothing had changed'.  Fortunately, a quick review of the application architecture followed by a discussion with the vendor and a network trace revealed the problem, but finding the right solution was far from simple. 

The system consisted of a conventional win-32 application which linked to a mainframe based system. There was nothing particularly special about the Windows application, but mainframe link was a little more interesting. Each application session required a unique IP Address/Port combination to work correctly. Obviously when running in a conventional single user workstation environment this wasn't going to be an issue. With only a single session per workstation and each workstation having it's own IP address the application would work just fine using its default port assignment, but in WTS based system multiple sessions exist on the same server so some means of assigning a unique port address per session was required.  The vendor hadn't designed the application with WTS in mind, there was no native means of checking for or negotiating a unique port assignment. The vendor did however offer an accommodation that would make the application work with WTS. The application could be forced to use a non-standard port number through a command line argument and the vendor's recommended solution was to  leverage this and give every user their own a personal port number. This is clearly supportable but not very scalable requiring a separate published app for each user. 

A better solution would be to implement a dynamic port management system running on the XenApp server to assign and track port numbers for each session, and as it happens this is just about what had been done.  Someone had developed a script using a TN3270 terminal emulator to negotiate per-session port assignment, and it this script that was no longer working. Remember I said 'nothing had changed', not true; the move from MetaFrame to XenApp also required a move from Windows NT to W2k3 and it was this OS upgrade that killed the script.

Digging any further into the root cause wasn't going to help us much. The script was  unreadable and undocumented, the script developer had long since left the company, and the terminal emulator vendor had gone out of business.  Lovely.

So where to now?

Initial review showed that there were plenty of different ways to tackle this one:

• Always at the top of my list is to ask the vendor to change the product. Unfortunately the vendor wasn't that interested; the customer had already made it clear that they would be retiring the app in a couple of years so didn't have much leverage and the existing solution appeared to work well enough for everyone else.
• We could have followed the vendor's recommended best practise and implemented a unique published app for each user. Aside from it being 'ugly' it had the advantage of being simple to implement and it was vendor best practise which always carries a lot of weight.  But it was discounted early on because it was thought to be too difficult to maintain with a user population that changed frequently. 
• Replacing or reworking the prior custom script would have worked, but neither the application owner, the vendor, or the in-house Citrix team were willing to take responsibility for developing and supporting the solution so this too fell by the wayside.
• The Citrix team were keen try the XenApp Virtual IP Address management feature to assign a unique IP address for each XenApp session and use a single default port address for each session.  This made a lot of sense;  no custom scripting was required and the vendor agreed that the proposal would work. But this was going to be a new and, at least as far as the IT engineering and operations teams were concerned, untested technology.  Just as significant, the application team weren't willing to pay for the Citrix team to design the solution and the project team were not willing to wait for the solution to be developed. 
• Next on the list came VDI. Even though it suffered from all the problems that killed the virtual IP proposal, there was sufficient interest from IT leadership to push the concept to the top of the pile. Unfortunately, but undeniably rightly, it was eventually decided that this was not the right time to trial VD technology and the idea was shelved.
• The last solution put forwards was to implement 'Single Session XenApp Servers' (What?).  A number of XenApp servers would be created as guests in the virtual infrastructure, each would be limited to a single session per server and treated them as any other XenApp server.  As wasteful as this sounds, it is by far the best solution on the table.  It is fully vendor supported, no new engineering time or user training is required. The project gets back on schedule, the user gets the app, costs are limited to a handful of W2k3 server OS licenses and as an added bonus, the VI environment is provided free of charge to the project as an incentive to move off physical servers.

What has this got to do with the VM Hosted Apps feature in XenApp 5.0 Feature Pack 2? Everything.  Forget for the moment that this implementation is built on a server OS and is  deliverable today; this is exactly why VM Hosted Apps (either as a Citrix XenApp feature, or as a general capability delivered by any other vendor) is so important to application delivery.  VM Hosted Apps provides a simple solution to reducing the complexity of application delivery across multiple client and hosting platforms.  

One app delivered seamlessly out of the box exactly where the user needs it, how the user expects it. Quickly, efficiently and above all else simply.  Behind the scenes it may be ugly, complex and far from efficient, but this is just today.  Come next year when VM Hosted Apps is a fully baked feature and hopefully renamed something more appropriately descriptive, it's going to be a very significant additional tool in the app delivery armory. 

Unless of course you are a Quest customer, if so I'm sure you have been doing this sort of thing for years.

XenApp Fail

Layering Citrix' management and presentation services on top of a real desktop OS was about the first serious technology suggestion I made to Citrix's leadership when I was offered a seat on their Customer Council.  That was about 7-8 years ago, when Citrix was still a single product company,  and well before there was any talk of desktop virtualization.  At the time I had a specific need that justified the suggestion, but it was an extreme use case with little likelihood of widespread adoption and required so many new technologies to make it a reality that it would not have made economic sense to Citrix to consider it at that time.  Moving on to today, Citrix recently announced VM Hosted Apps for XenApp that will be shiupped as part of XenApp 5.0 Feature Pack 3. VM Hosted Apps is the logical progression of that original concept.  It's been shaped a little here and there by evolutionary pressures from newer products and the opportunities afforded by virtualization technologies, but it's still very close to the original idea.  So with that background I was rather enthusiastic when I heard it was coming.  But now that it has been officially announced I'm having a hard time sustaining my enthusiasm.  

The biggest problem I have is that it's still incomplete. Given Citrix's position in the market I would have expected that they would have been able to afford the time to develop and release a fully baked set of product features rather than the rather confusing and somewhat incomplete offering that has been announced.  

For starters is it a XenApp feature or a separate product.  In order to implement VM Hosted Apps you have to build a separate farm and for me at least if it requires a separate farm then it fits more closely with the notion that it isn't (yet) a XenApp feature, more a separate product.

Another hint at a possible rush to market is the one to one host OS to published application limitation.  XenApp has had session sharing for as long as I can remember, so there's absolutely no reason not to expect this capability from any new Citrix delivered application publishing solution. Dropping session sharing from the initial release of VM Hosted Apps can only be because of a need to ship before the session broker was fully baked.  

Expanding on the separateness theme, Citrix has had endless criticism over it's inability to deliver a unified management console since Presentation Server 4.0 introduced the AMC and I don't think it is going to end here. Delivering a new XenApp feature that requires a separate MMC snap-in isn't the most obvious way to tell people that this criticism is being addressed (it is being addressed but that's beside the point), especially after the promise that the old CMC is finally being retired.

Then there's the name; 'VM Hosted Apps' sounds like an internal place holder though up by someone in engineering rather than the typical marketing lead feature name we have come to expect from Citrix. OK, we all criticize marketing for interfering with familiar product names, but for the most part the pain of transition is worth the end result. The real problem I have with VM Hosted Apps is that it has neither the stickiness of a good marketing name nor the accuracy of a well thought out engineering lead name.  

What makes this feature worthwhile is not that it hosts apps on virtual machines (not to mention old-school physical machines), but that it delivers apps running in seamless windows from a desktop OS.  Why hide this benefit behind an obscure name that could describe any app hosting technology is beyond me.  Come on guys, can't you come up with one of those marketing department generated, focus group blessed psuedo-names – 'Tru-Windows for XenApp', 'XenApp Grid' or even 'AppCloud', anything but 'VM Hosted Apps'.  Still, it's not all bad, given Citrix's record we can expect a new name to be announced in San Francisco next year.

There's no denying that this is exactly what Citrix needs to round out their hosted application delivery product line and it sets up even more convergence opportunities for the future, but as a finished product it falls well short of the mark.

Citrix Project Independence Client Hypervisor - early alpha review

Gabrie van Zanten has posted a fine review of an early beta alpha of Citrix's Project Independence Client Hypervisor (this is the real Type I Hypervisor, not the mythical XenWorkstation Type II Hypervisor). Read it at his blog www.GabesVirtualWorld.com. I've not had chance to take a hands on look at myself yet, but after reading Gabe's review I don't feel the urgency any more. 

 I've spend a fair amount of time since the initial Project Independence announcement thinking through what an idealized client hypervisor implementation should look like. Fairly high up on the feature list was the ability to install multiple guests and treat them as separate layered personalities with 'firewalled' links between them through which applications and other services might be allowed access. The simplest implementation might be the ability to work on a corporate Windows application running in a secure guest VM, while maybe talking to a friend via Apple's iChat, but without the ability to cut data from the corporate app and paste it into the iChat window.  Really just the same sort of thing I can do today with XenApp, but without the network tether. 

 I've already managed to convince myself that this could be a killer feature, so I was very happy to read that Citrix have already implemented this using ICA as a means of establishing links between the primary desktop and applications running in a guest OS. OK it sounds like there's no support for anything but Windows as the primary partition yet, so the detail of my hypothetical scenario might not work out today, but you get the general idea.

Gabe has some good points about the amount of memory the system will require; it certainly looks like you will need a powerful laptop to get the best out of the product as it stands today.  Still maybe this is all part of the Obama stimulus plan, it’s time to throw away those netbooks and buy yourself a real laptop. 

Granted there's an ocean of possibilities between an alpha system and the production release, but it sounds like Citrix are heading in the right direction.

Citrix blur the lines between app and workload delivery - I

Citrix have always had a to face the challenge presented by having to stay a couple of steps ahead of Microsoft's Terminal Services feature set and more recently by competitors looking to steal parts of Citrix's core customer base.  So after making it clear to VMware yesterday that they have a fight on their hands for the data center and cloud virtualization markets, Citrix went on to make it clear that they haven't forgotten their base either by announcing Feature Pack for XenApp 5 and blurring the lines between application delivery and workload delivery.

Sridhar Mullapudi and Vinny Sosa blogged about the change yesterday, and provided the following chart highlighting the changes 

	Advanced	Enterprise	Platinum
Citrix XenServer	X	X	X
Application streaming	X	X	X
EasyCall voice services	X	X	X
Workflow Studio orchestration	X	X	X
Load testing services		X	X
Profile management		X	X
Provisioning services			X
Single sign-on			X

It should be obvious to everyone by now that XenServer is not really a part of XenApp, so it's arguable that it should not be included here.  However, there is one key difference that Citrix have made that might not be immediately visible.  By making XenServer a part of XenApp, it means that XenServer is now supported by Citrix as part of any existing XenApp support contract, so even the smallest of XenApp users can now implement XenServer as a fully supported production platform for no extra cost. This is huge in itself, and is another direct attack on VMware who have been making a play for hosting XenApp on ESX for some time. OK, it should be noted that officially this support status only applies to using XenServer to host a XenApp workload, but if ever a rule was there to be broken this one is. 

Moving App Streaming down to XenApp Advanced is a strong way to shore up sales of Advanced to make it attractive for customers to stay with Citrix rather than jump ship to cheaper products or forego their Subscription Advantage agreement at a time when IT budgets are under pressure. The same could be said for the other major additions, Workflow Studio, Load Testing Services and Provisioning Server. Citrix came under substantial criticism last year for a perceived lack of new features in XenApp 5.0, this update should go along way to win back customer loyalty. 

It leaves me wondering though, if Citrix are in a position to announce this many changes before VMworld Europe.  What will we see at Synergy in May?

More later.

Moving XenApp into the cloud - is Amazon growing a second Long Tail

Chris Fleck (Citrix VP of Solutions Development) recently published a short guide showing how to build XenApp Servers in Amazon's EC2 cloud using a Community AMI (Amazon Machine Image) that allows anyone to be up and running with their own XenApp server in minutes. In releasing XenApp to the cloud Chris has given every IT professional in the world access to their own personal XenApp lab and by extension has given every potential XenApp customer or Citrix Ready Partner a very low cost, risk free means of assessing how their applications could be delivered by XenApp (not a bad day's work Chris). 

Of course, this opportunity isn’t limited to Citrix, I would expect Ericom and all the other Remote Desktop Services vendors to follow Citrix’s lead and provide access to their products this way. If nothing else this will enable potential customers to perform side-by-side comparisons between competing technologies far more readily than is possible today. 

I wonder how long it will be before companies like Alpiron (see my previous posts on data center power management for more detail on Alpiron) release their own Community AMI to allow potential customers to evaluate Alpiron’s server power management system running in EC2. Granted there's nothing to stop Alpiron from doing this today with their own servers but it makes a lot of sense for a company selling power management solutions to be able to bolster their Green credentials by placing their demo environment in one of the most power efficient data centers in the world. 

Extending this line of thought a little further, what else looks like a good candidate for the cloud. At this point, I’m willing to suggest that almost anything without an SLA could be placed in the cloud. I still hesitate to suggest considering production environments. Cloud technologies are still immature and as such subject to greater threat from unidentified risk than more mature technologies would be; this alone is enough for me to want to steer clear of suggesting that production systems are good candidates for the cloud. 

One of the more interesting possibilities that application delivery from the cloud brings is in the realm of business continuity services.  About 10 years ago I spent a year or so working for a systems integrator that worked primarily with small businesses.  Most of these companies’ disaster recovery plans were limited to off-site storage of their backup tapes and an understanding that they might need to purchase replacement servers in the event anything untoward happened and hope that the backup tapes had worked. Skip forwards to the day after tomorrow, and that same company could have a cloud based DR plan that required them to mirror their data to the cloud every night, and in the event that disaster were to strike their recovery action would be to launch however many private AMIs (copies of their physical IT infrastructure) and they could have their entire IT infrastructure back up and running in the cloud in a couple of minutes.  This kind of warm standby service has always been beyond the reach of all but the largest IT shops, but with the advent of cloud services just as eveyone can afford their own lab environment, now every business can afford warm standby.

Which got me to thinking; Amazon has always been about the Long Tail, making most of their money by selling the products that bricks and mortar stores can’t afford the shelf space to stock.  It’s not beyond the bounds of possibility that with EC2 Amazon is growing a second tail, providing a low cost framework for DR and Business Continuity services at a price that traditional DR service providers can’t afford to match.