For those that are paying attention on the happenings in the hypervisor world - Oracle launched OVM 3.0, it's next generation hypervisor last week. One of the key messages as it relates to hypervisor that we have been highlighting is that virtualization needs to be purposeful. For virtualization to be purposeful, it needs to really understand the workloads that run on top of it. If the hypervisor acts as just a black box, then from a performance perspective, you just lost a few ticks of your CPU without getting much for it. On the other hand,if the hypervisor is workload aware or "application aware" then you get the following benefits
- Better performance as the application is better tuned for the hypervisor
- Better reliability during failovers/live migrations etc - less chance of application data loss/application config losses.
- Rapid buildout of your virtual environment while keeping less gold images ( vm image management is a pain for those that have done it) as you use templates to build things out - keeping things modular in nature.
- Allows change in configurations without having to completely rebuild and store new gold images and takes away worries about versioning etc.
- Allows the user flexibility to manage patching activity much faster -rather than an all or nothing approach..
Interestingly enough at Loudcloud and later at Opsware, we had the exact same viewpoint of how application infrastructure needs to be built out in a physical environment (at that time). Our approach in Opsware's Server Automation platform was template driven against BMC's Bladelogic vision of managing golden images for each and every minor variation.
I will try to add more later on the compare/contrast of template driven buildout of infrastructure against a image/snapshot driven buildout.