Having entered the virtualization space late in the game, Microsoft is working hard to recover lost time.
So far, the Redmond giant's strategy has had three key moves. First, the company announced the Windows Server Virtualization hypervisor and System Center Virtual Machine Manager tool, two competitive products it hopes will challenge VMware Inc. for market leadership. Then, it acquired Softricity Inc., a recognized leader in application virtualization with its SoftGrid product. Finally, it entered into an agreement with XenSource Inc. for virtual machine interoperability between its hypervisor and the most famous open source virtualization platform, Xen.
This last step raised a lot of hackles in the virtualization industry, with reactions from many vendors, including VMware. Customers found themselves faced with uncertainty regarding conflicting information, the complexity of the migrations and the impact on business. Summarizing the key points of any problem always helps to clarify the issues, so here are the key elements in the Microsoft-Xen situation:
- Microsoft has been involved in the Xen project since the very beginning. SWsoft Inc. reports Microsoft's upcoming Windows Server Virtualization and Xen hypervisors have a similar architecture.
- Xen is open source.
- Microsoft Windows Server Virtualization will be out within two years.
- VMware has been proposing a virtualization standard for more than a year, but the work has severely slowed down because of differences of opinion, particularly with regard to the XenSource approach.
- VMware rarely comments on competitors' moves in the virtualization market, but in less than a day it reacted aggressively against the Microsoft-XenSource agreement.
- Xen is adopting a paravirtualization approach, involving guest operating system kernel modification, permitted only on open source platforms.
- With the new features inside AMD (SVM) and Intel (VT) processors, users can achieve the same capability to run operating systems with paravirtualization. In the future, new generations of this technology will be able to achieve the same performances.
- Virtual Iron Software Inc. publicly announced that its products would be based on Xen, but that the company will dismiss the paravirtualization approach.
- Novell released its new operating system with Xen but limited support to its own operating system for virtual machines.
- After the Microsoft-XenSource announcement and Novell operating system release, Red Hat Inc. suddenly warned that Xen is not ready for production, despite the big announcement of a comprehensive virtualization strategy on March.
Given these points, customers looking at Xen and Windows Server Virtualization as a viable alternative to the existing VMware ESX Server solution have now to ask themselves:
- Why key Xen supporters are offering limited support for it even after implementing the hypervisor on their commercial solutions.
- Why Microsoft didn't opt to participate in VMware and Xen hypervisor standardization works instead of working on an interoperability solution.
- Why Microsoft needed XenSource to implement Xen interoperability when it knows the project so well.
- Given licensing and support issues, why moving a native virtual machine from its hypervisor to another should be better than committing a physical-to-virtual (P2V) migration.
Answering those questions leads to some notable conclusions. First of all, the recently announced Microsoft-XenSource deal has value purely as a political pact. Considering that it's impossible for XenSource to benefit from Microsoft's know-how because of licensing incompatibilities, the agreement is unilateral, providing technical benefits to Microsoft only -- if there are any. The main goal will be to verify if a XenEnterprise Linux virtual machine is as fast as, or faster than, the same machine built from scratch on Windows Server Virtualization. This will be impossible until both companies release their products, assuming hypervisors will feature the interoperability patch at that time.
Another important conclusion is that Xen is a project with high potential but an uncertain destiny. In the immediate future, customers who decide to adopt it even through the Novell or Red Hat operating systems, will live in a half-supported world where they cannot use the hypervisor for anything other than virtualizing the same platform. In this scenario, OS partitioning products like SWsoft Virtuozzo are a better fit.
In the long run, assuming the next-generation processors will be able to provide the same performance paravirtualization offers today, Xen will lose its technical advantage and be obliged to compete au pair with other products. At that point, customers will need a compelling reason to adopt the open source hypervisor, without the experience of VMware or the support capabilities of Microsoft.