OK thanks for the link. It makes it clear that you need to disable I/O metrics within SDRS. As far as the provisioning of the LUNs, the storage team will then just provision a set of identical LUNs, and let FAST determine how to provide the right performance? Do I still need to tell the storage team which datastores have VMs with high performance needs, or is that unnecessary now with FAST since it detects this on the fly and makes adjustments based on performance characteristics?
If defining LUNs with different storage characteristics is unnecessary with FAST, then the whole idea of profile-driven storage would seem unneccessary. If every LUN is on FAST, does that make profile driven storage unnecessary?
Also, previously when LUNs predictibly gave a certain number of IOPs, and certain throughput and latency, I could measure the requirements of my VMs for these metrics, and then plan the placement of my vms depending on how much performance could be delivered by a LUN. Now, since there are no hard fast numbers that a LUN will deliver, but LUNs deliver performance as it is needed, how do I know when I will run out of available IOPs, latency, throughput, etc? Does all of this consideration get moved to the storage layer and off of the virtualization layer and entrusted to FAST? How do we manage this situation as VMware admins?