One of the interesting questions in the virtualization industry is why desktop virtualization has not achieved more traction more quickly. While there have been a growing set of pilots since the start of 2008, only a small proportion of these pilots have gone into production. Furthermore, even for the ones that have gone into production, they have not been followed up with any kind of large scale migration of more users into the VDI environment. Note that these are generalizations and that there certainly are some very large enterprises that have significant VDI implementations in production.
David Bieneman, the founder of VizionCore has joined up with J. Tyler Rohrer formerly of the VMware VDI team and the team from VMsight lead by Jonathan Alexander to create a new company – Liquidware Labs. Liquidware Labs is focuses upon two aspects of the problem listed above. The first aspect is to define a methodology (Assess–>Design–>Deploy–>Manage) for VARs and service providers to use in VDI projects, and the second is to provide a toolkit that implements much of this methodology in software. The toolkit is based upon the vmSight product which had the following interesting and unique attributes:
- An agent was available that could be installed in a VDI Guest, and provided unique guest identification via a patented feature called “Connector ID”. This same agent collects metrics from within the guest.
- A virtual appliance is installed in each host that collects data from the virtual mirror port on the virtual switch within the host. This is the exact same method of data collection that the forthcoming AppSpeed (acquired when VMware bought B-hive) product uses. One important difference is that the vmSight virtual appliance worked in Hyper-V and Citrix Xen environments as well, making it the only cross-platform monitoring product that collected data in this manner.
- A monitoring station collects data from the two sources above, and also directly from the vCenter API’s and is also packaged up as a virtual appliance.
With the forming of LiquidWare Labs, the vmSight product has been rebranded as Stratusphere, and has been focused upon the problem of providing VAR’s and service providers a toolkit that implements the Assess–>Design–>Deploy–>Manage methodology in a software solution that assists these VAR’s and service providers in the process of assessing which users should be the subject of a transition to VDI.
As a part of focusing the technology upon this new problem, Liquidware has created two new composite metrics that should be immensely valuable to partners who are in the assessment stage of VDI projects. Those two new metrics are VDI FIT and VDI UX. It is important to note that both of these metrics are composite in the sense that they are composed of multiple other metrics and then run through an algorithm to create the new composite metric.
VDI FIT is designed to be used on existing physical desktops to determine which of those desktops are and are not good candidates for a migration to VDI. VDI FIT is comprised of the following metrics:
- System CPU Load
- User CPU Load
- Memory Load
- Disk Load
- Network Load
- Network Latency
- Graphics Intensity
- Use of Known Bad Applications
These metrics are synthesized into a Good, Fair or Poor rating for each physical desktop. By quickly and easily identifying the users and desktops that are great candidates for a VDI migration, the service provider can get a fast win for the first phase of the VDI project by focusing upon low hanging fruit (which is exactly how server virtualization got its start with servers that were low hanging fruit). By identifying users who are poor candidates for migration, the project can avoid embarrassing failures and political heat from users, things that can quickly derail or stall these projects.
VDI UX (User Experience)
VDI UX is about monitoring and understanding the experience of the user once their desktop has been virtualized. This is an extremely ambitious and important undertaking, since measuring user experience for VDI hosted applications is every bit as difficult as doing this for Citrix/XenApp published desktops was and continues to be. There are many challenges here, but most of them are rooted in the fact that the notion of end user response time within applications is obscured by the ICA and RDP protocols. There it is critical to understand that this is a problem that has no perfect solution, and that a good solution that is a reasonable proxy for actual end user experience is all that is possible. This is VDI UX is designed to provide. VDI UX is comprised of the following metrics:
- User Login Durations
- CPU Queue Length
- Memory Page Faults
- Disk Queue Length
- Failed Network Connections
- Network Latency
- Graphics Intensity
- Application Load Times
- Non-responding Applications
The combination of the Liquidware methodology with the Stratusphere tool kit and these new metrics creates a tool kit focused upon providing VAR’s doing VDI projects a consistent set of processes and metrics by which to measure these projects. This will have the effect of getting more successful VDI projects into production more quickly which is a good thing for VAR’s, their customers, and the vendors that provide VDI infrastructure technology. David Biemenman’s background as the CEO of a VAR focused vendor (with VizionCore) and Tyler Roher background with Foedus means that the company “gets” the VAR channel, how it works, and how to make money in cooperation with it. This is something that very few vendors understand as viscerally and completely as does Liquidware which bodes well for the success of the company with the channel.
From a product and technology maturity perspective there is also some good news. The fact that the Stratusphere technology is based upon the vmSight product which has many releases and QA cycles behind it (and some deployments in production VDI environments) means that most of the immaturity speed bumps that occur with a new product will likely be avoided. That said there are significant new problems that Stratusphere is trying to address that are beyond the scope of what vmSight tried to do, so there will doubtless be some issues as the set of functionality expands with new code designed to address these opportunities.
Finally, there is the question of how to measure and assure the performance of a VDI implementation once it is in production. This is something that the VDI UX metric is designed to address (under the assumption that the customer chooses to license Stratusphere for ongoing use after the project is complete). However, as was mentioned earlier, there is still a great deal of work that needs to be done on this front before it becomes possible to measure real end user experience with the same kind of fidelity and granularity in VDI implementations as is the case with existing fat client desktop implementations.