logo

My reaction to this fuss is that it's a lot of misdirected energy. While performance is always important, think about the real reasons you deploy a hypervisor: greater CPU utilization, more agility, rapid deployment... When you make a strategic decision to deploy a hypervisor, you have to take the good with the bad. For benefits like better utilization and agility, you have to sacrifice on things like performance, security and complexity. When I say sacrifice, I mean that hypervisors will never perform the same as native systems. They will have lower security (open zoning your LUNs to support vmotion for example), and they will add complexity from a management perspective. However, in many cases, these are trade offs worth making for some customers. So if performance isn't the main place to focus, what is? To me it's always been about providing the benefits of hypervisors - utilization, agility, rapid deployment, etc. - while removing as many of the "penalties" as possible. Egenera vBlade solves many of these issues. It solves the security issues with a unique layer of I/O virtualization delivered by the PAN architecture. It solves the complexity issue by integrating the management of virtual machines right into PAN Manager, allowing the user to manage virtual and physical servers in exactly the same fashion, from one location. vBlade won't change the performance characteristics, as these are specifically tied to adding a hypervisor into the mix. So, maybe the focus shouldn't be on detailed performance metrics as the main selling point. Besides, we all know performance data is subjective at best. But perhaps the industry should be focused on solving real enterprise problems through augmenting the strengths of hypervisors with the strengths of enterprise-quality management systems.