With all the news surrounding the virtualisation capabilities provided by Microsoft Windows Server 2008, it seems like a good time to take a look at the capabilities offered by competing environments.
Open-source suppliers have been including virtualisation technologies in their Linux distributions for several years now, with commercially supported offerings appearing in late 2006. These offerings have reached a level of maturity and functionality, offering features such live migration and resource control (the ability to dynamically add and remove CPUs and memory to a virtual server). For example, Red Hat's Enterprise Linux Advanced Platform product supports both Red Hat Enterprise Linux and Microsoft Windows guests. While Windows guests need to be licensed according to Microsoft practices, an unlimited number of Enterprise Linux guests can be hosted for no additional charge.
So today's customer, finding that there is a growing selection of virtualisation products on the market, is faced with the decision of choosing the best solution for his or her needs. Perhaps the first realisation is that the future is going to be "multi-hypervisor", with different virtualisation technologies being used in different deployments, in exactly the same way that enterprise and SME customers have heterogeneous operating-system environments. And, undoubtedly, Linux-based, open-source virtualisation will be appropriate for many environments.
Given a multi-hypervisor future, it becomes necessary to take a rather more global view of the impact of virtualisation on a modern IT infrastructure. At the top of the list of concerns, as customers will quickly appreciate, will be virtualisation management and deployment flexibility.
Every customer, big or small, will shy away from any solution that complicates system management. Virtualisation, with invisible servers moving dynamically around the physical infrastructure, can do that in spades. Meanwhile, system-management application vendors are faced with either making their products support a multi-hypervisor world or limiting their scope, thereby forcing customers to use multiple management tools. So, the availability of a common virtualisation-management fabric for the multi-hypervisor world is going to be high on every chief information officer's wish list. The solution to this problem is for the industry to provide an open system-management standard for virtualised environments.
The open-source "libvirt" project provides a standardised management interface that neatly isolates the hypervisor from the management application. Libvirt has been steadily garnering adherents, including Red Hat, IBM, Sun, Novell, Ubuntu and others. While libvirt represents a threat to virtualisation vendors who provide proprietary system-management tools, it offers customers the opportunity to manage their diverse virtualisation environments in a consistent manner. If industry vendors continue to join and collaborate to drive the libvirt standard and capabilities, customers will then be free to choose the most appropriate virtualisation technology — all deployed under a consistent management environment.
One of the promises of virtualisation is deployment flexibility: the ability to quickly provision a server using existing resources. But, as usual, the devil is in the details. Today, a chief information officer can choose to deploy applications in several ways: on a traditional ("bare metal") server, on a virtual server, in a compute cloud (such as Amazon EC2) or perhaps as part of an appliance. The applications must be able to run in any of these environments as though they were all the same. It may sound trite, but the ability to run "any app, anywhere, any time" is what today's chief information officer is looking for.
Open-source virtualisation capabilities are already mature, well proven and highly cost-effective. The open-source community is moving to the next level to provide incredibly flexible environments that are genuinely manageable, regardless of the deployment model or underlying technology.