logo

Is this a good thing? In general, sure. We've been throwing money at barely used servers for decades. Virtual machines will enable the customer to get more out of the capital invested in servers. That's great for the customers.... Not so great for the hardware vendors. And while higher utilization rates will make customers happy, it will also lead to the next great data center challenge managing all of these virtual machines, which of course leads us to the next technology surge....Companies who "manage virtual servers." This push to manage virtual machines will be chaotic and will no doubt induce risk into the data center. In order to reduce or remove the risk, the problem must be looked at from a different perspective. The underlying infrastructure of these virtual machines must be resilient. This doesn't imply a hardware solution as much as a software solution. What happens if a hypervisor fails? Prior to virtual machines, if a server failed it was catastrophic for a single operating system and typically a single application. In the new world, when a hypervisor (the new hardware "platform") fails, it can be catastrophic for several operating systems and applications, creating a larger impact for the data center. More subtly, what happens if an IO path fails or a NIC fails below the hypervisor? Component failure can now cause much more damage then in a single server model. So what's the answer? Resilient Virtualization. The hypervisor is an important tool, but alone it is not the answer. It needs to be surrounded by capabilities that make it safe for enterprise use. Failover, Disaster Recovery, IO multipathing, load balancing, etc. are just some of the enterprise capabilities that are necessary to ensure that the hypervisor can live up to it's hype. Putting a GUI on the hypervisor and "managing virtual machines" might add more risk to the data center today. Adding enterprise capabilities to really manage the virtual machines and make them safe for the enterprise is what is needed.