Cloud computing challenges many long-standing paradigms of the PC computer era. In my last post, I discussed the technical challenges facing the OVF standard in virtual system portability. Even if we assume that OVF will overcome the technical barriers, an even larger one looms: licensing.
The dominant licensing model for PC software over the last 25 years centers on hardware systems. Whether a vendor strictly enforces their license or not, the EULA for desktop software typically permits installation and use on a single computer. The paradigm example of this remains the Windows and Office products from Microsoft – the license key is tied to various properties of the hardware through “product activation”.
In response to the complex problems this model presents to businesses, Microsoft offers the volume licensing paradigm. Even in this model, the license counts are still tied to hardware characteristics. The most obvious example is the licensing of Windows Server by number of processors on a target system. However, in a cloud infrastructure, the underlying hardware is inherently shared with other tenant virtual systems. It becomes impossible to accurately determine the actual processing power available to a given virtual system. Volume licensing was never conceived as a virtual system licensing scheme, nor has Microsoft ever claimed it as an appropriate solution.
The applicability of traditional licensing models to cloud infrastructures is questionable at best. An obvious limitation it creates for OVF is its use in software distribution. Imagine a world where most ISV’s distribute their data center software as ready-made, bootable OVF images. When it comes time to update or upgrade systems, customers can apply updates or simply migrate to a new self-contained and fully validated image. The cost savings across the industry to ISV’s and customers avoiding costly installation and basic configuration steps would probably be on the order of billions of dollars over the typical software lifecycle.
Such a world cannot exist today for Windows-centric data centers or application developers. The only redistributable Windows OS is the trial version that cannot be converted to a permanent license. If Microsoft supported a “bring your own license” (BYOL) model, such images could be converted to permanent installation status. Vendors could then leverage the vendor-neutral OVF standard to distribute turn-key products.
More generally, licensing requirements still impede adoption of Virtual Desktop Infrastructure (VDI) built around OVF to host employee desktops as virtual systems in the data center. IT departments are starting to consider the cost and security advantages of VDI. Citrix is starting to see more substatial uptake with our XenDesktop offering in this area. However, broader adoption of VDI is impeded by the cost of licenses. Part of the cost for Windows is that Microsoft requires the special subscription Virtual Enterprise Centralized Desktop (VECD) license on top of other costs. The situation for Mac OS X is more dire – Apple simply doesn’t support virtualization of the desktop OS.
The existence of legacy licensing models that use dongles, MAC addresses, processor identifiers, and the like are less common today, but we still see customers foiled by such problems. This is particularly true when it comes to migration of existing systems into the cloud. However, even if most ISV’s change EULA’s and licensing schemes to officially support a cloud-based world, the optimal use of OVF as a tool of the IT cloud will remain impeded by the realities of how the OS components are licensed.
How do these licensing concerns impact your use of virtualization technologies? Have you encountered licensing problems trying to migrate to a cloud platform? Would you use a ready-made OVF images if Citrix or other vendors made them available?