Dealing with the Volatility of Virtualization
|
One of the more troubling aspects of deploying virtual machines is the percentage of them that become unstable over time. Current estimates suggest that anywhere between 10 to 15 percent of virtual machines become unstable. Given the relative infancy of virtualization it's hard to exactly quantify how many virtual machines are actually unstable. But it's fair to say that the rush to deploy virtual machines creates opportunities for lots of mistakes largely because people will take short cuts when it comes to application testing. This creates a major challenge for IT managers charged with keeping the mass movement towards virtualization from devolving into chaotic mass of virtual machines that nobody is keeping track of, or worse yet, doing anything to fix because they can't identify which virtual machine deployments are unstable. This is potentially a very big problem because IT organizations tend to replicate virtual machines all over the company once they initially get it up and running. So if it turns out that at later date that virtual machine becomes unstable, chances are it's not going to be confined to one isolated incident. These are the types of practical issues that providers of systems management tools are trying to deal with. For example, the V-Commander management tools from Embotics provide tools for tagging and tracking virtual machines. It can also be used to identify rogue virtual machines that have been deployed without authorization by comparing and contrasting all the virtual machines in the environment with the list of virtual machines found in the Virtual Commander repository. In its latest release Embotics has added policy management capabilities that make it easier to manage virtual machines deployed across different types of zones while also discovering the lineage of a virtual machine so problems that appear on one deployment can be preemptively dealt on all the deployments leveraging that instance of a virtual machine. It also provides better tools for quantifying the costs associated with deploying virtual machines and helping figure out which virtual machines are actually associated with what applications. As a cottage industry starts to emerge around virtual machine management, it's becoming pretty clear that there's a lot that can go wrong with virtualization. And given the fact that most IT organizations are just now getting started with virtualization, you can't help but wonder how much of problem IT organizations are going to have in the not so distant future when a lot more of the early deployments of virtual machines become increasingly unstable over time. |

Comments (1)
It seems that many virtualization managegement vendors are appearing, however, I don't see many focusing specifically on performance management for mission critical applications and business services that have both virtual and physical components. The complex interactions between tiers can be difficult to troubleshoot in purely physical environments, but add, for example, a virtual web server tier and you've brought the complexity up an order of magnitude. IT Operations is going to need new solutions (such as my company's Integrien Alive) to handle these environments that can use analytics to automate the massive manual efforts that go into troubleshooting with static threshold-based monitoring solutions.
Posted by Steve Henning | July 2, 2008 8:17 PM