virtualization.info

Yesterday, during its Synergy 2010 conference in San Francisco (see virtualization.info coverage), Citrix announced the availability of XenClient: the free client hypervisor formerly known as project Independence.

As published yesterday, the XenClient build released yesterday is Release Candidate.
A lot has been said about the huge delay that both Citrix and VMware accumulated to launch their client hypervisors, and the reasons behind it.
Despite that, Citrix managed to release a pretty stable version of the platform before VMware could do the same with it Client Virtualization Platform (CVP), and managed to support nine average laptops (three from HP, three from Lenovo and five from Dell).

After the keynote, virtualization.info sit with Peter Blum, Director Product Management and Marketing for this new platform, who detailed a lot of aspects never published before on XenClient.

The first and most important information to know about XenClient is that it doesn’t require Intel vPro.
A lot has been said about this topic too, and both Citrix and VMware have been criticized for the choice to have vPro as a mandatory requirement to run their client hypervisors, when much smaller competitors like Virtual Computer and Neocleus don’t need it.
Well, at least for XenClient, Intel vPro is highly recommended but not a mandatory requirement.

At today vPro is an umbrella for multiple Intel technology: VT-x, VT-d, Active Management Technology (AMT) and Trusted Execution Technology (TXT).
In its current form, XenClient only requires VT-x and VT-d.
The reason why Citrix is highly recommending to have a laptop certified for all Intel vPro technologies, is that in future XenClient may leverage the other two, like TXT to recognize hypervisor’s tampering attempts.

To guarantee remarkable guest OS performance, Citrix is not just leveraging VT-d. The company is using para-virtualized drivers for the top three Windows guest OSes: XP, Vista and 7.
At the moment, there’s no official support for Linux guest OSes, but the company reports that it works.

There will be several editions of XenClient. The current one is called Express and integrates the Citrix Receiver and another, new component called Synchronizer, which is needed to remotely provision virtual desktops and security policies on the hypervisor.

While Synchronizer takes care of virtual desktop provisioning, there’s not a specific solution to provision the XenClient platform itself on bare metal laptops. Anyway the product features a PXE subsystem and thus it can be used with any third party enterprise distribution product, including the Citrix Provisioning Server.

Citrix tentatively expects to release the GA of XenClient in Q3 2010.
The product is and will remain free and (mostly) open source. Source code is available already now but, like it happens for XenServer, it doesn’t include some proprietary components, like the user interface to switch from one virtual desktop to another.

As already said in our previous articles, XenClient is based on the same virtualization engine of XenServer, which is based on Xen.
At the moment XenClient is built on top of Xen 3.x. The new Xen 4.0 engine will be implemented as soon as it matures a little in versions 4.x.

Contrary to what the press reported on the reasons behind such delay (XenClient was originally expected for the end of 2009), the biggest technical challenge is not to support the myriad of laptop configurations on the market.  The complexity is in other aspects.

First of all, Citrix has to guarantee a properly balanced power management in XenClient, avoiding that the new hardware virtualization layer impacts too much on battery life or that it prevents power saving capabilities that modern operating systems offer today. To do so the company has to go through a complex fine tuning process and, while Intel is assisting, finding the optimal configuration for different laptop models is a time-consuming activity.
The second issue is related to support the graphic subset in a proper way: Citrix is well aware that a smooth visual experience is critical to guarantee the success of this platform.
The third, most relevant issue is to guarantee a comprehensive USB support, which apparently is rather complex to achieve.

Despite it may not delay the product release as the issues above, Citrix has yet another challenge, which is to develop a valuable user interface that doesn’t confuse customers (what has to happen inside each virtual desktop when I press the power off button of the laptop chassis?).

Citrix is aware of challenge and is investing big in usability. The company has a dedicated team that is solely focused on this and researched around user interfaces for a few years.
More than that, in March 2009 Citrix hired the former Director of User Experience at Salesforce, Catherine Courage, who now is Vice President of User Experience, and is now leading the company’s effort in improving usability across all products.
Even during the Synergy conference, Citrix is even inviting attendees to sit down and use XenClient to study how the interaction works and how it can be improved.

The challenge for the usability team is big not just because of the inherent complexity of the XenClient approach, but also because Citrix is not targeting the average consumer with this product and thus some design choices will impact the user experience in any case.
XenClient has been designed to appeal corporations and IT professionals, so there’s a major focus on security.
For instance, the most typical operations between virtual desktops that standard hosted desktop virtualization solutions permit, like copy & paste or shared folders, are denied.
Despite Citrix is using a seamless window approach to show the two virtual desktops’ applications side by side, the only way to exchange data between them is to go through the network.

What is the long term strategy of Citrix for this product?

Of course XenClient is yet another container where to stream applications, which is where the company makes money, but Citrix is looking at additional opportunities for its client hypervisor.

The product has been designed to be extensible. At the first release there will be no APIs, but Citrix envisions a 1.x version that will include interfaces and an SDK, and that will be leveraged by partners to do interesting things, like enterprise management.
The idea is that XenClient may become a de fact standard platform on which 3rd parties can add value with different plug-ins, which seems exactly the vision that Phoenix Technologies had for its HyperCore client hypervisor (based on Xen too) that never became a success.

The vendors that already offer a client hypervisor may see this a great opportunity to lower their R&D costs while staying competitive.
Virtual Computer for example
didn’t exclude the idea that one day XenClient could replace its existing client hypervisor (which is based on Xen as well): the only problem of course is that XenClient is not yet 1.0 and Citrix has a lot of work to catch up in terms of hardware support and capabilities.
If Citrix will succeed, considering that Virtual Computer is not monetizing on the hypervisor itself, there’s no reason to not offload most of the hypervisor’s development costs to Citrix, while focusing on innovation somewhere else. Also because Citrix is an investor of Virtual Computer.

Citrix is not in hurry: XenClient seems a long term project and the company sees this first release as a first exploration of the market opportunities. Subsequent major releases will be more solid and feature-rich to appeal enterprise customers, and over the long term, depending on the market trends, Citrix may even consider a porting on other architectures (ARM-based tablets anyone?).