Not sure if you came across this post, but the VMware team describes how the architecture is different than the Xen/Hyper-V design. These sort of design differences are also prevalent in your storage devices. As I am sure many of you are aware, the management firmware/OSes of most storage systems is based on a commodity OS: Windows, Linux or Unix. Some systems deploy a simple iSCSI target on top of Linux and call it an appliance. Other vendors redo the networking stack in order to optimize it for iSCSI traffic. Others rewrite the OS from the ground up to handle storage traffic. How the iSCSI target is implemented defines the approach, features and redundancy model of the system. Some vendors, like StoneFly take a layered approach to "storage virtualization ." Separating the storage intelligence from the physical hardware. Other simpler iSCSI implementations actual tie the physical resources to the logical ones; the advanced functionality is then dependant on the physical devices.
When you are planning your infrastructure, be sure the level of redundancy you require is considered.
What layer does stuff happen: Data processing, traffic routing, RAID, volume creation, storage management? Will the implementation introduce overhead in to your real life performance. Quiz your vendors to make sure you are clear on what you are getting.
