One thing that I just thought of that you didn't mention in your article is that you can also use your redundant paths to the SAN to balance your IO by setting the preferred path for specific LUN's to the secondary path in the event that you have one path become saturated. This can be done in the Virtual Infrastructure manager, and you can use putty sessions to the ESX host to monitor the true IO metrics for the HBA's to know if they are queuing commands heavily or not, so the old adage that you can't track down bottlenecks in a virtual environment no longer apply as they used to.