I am trying to compare VMware's new VSA with the HP LeftHand VSA. Both products allow you to run the SAN node as a VM. Reading VMware's documentation I get the impression that VMs on the same host as the SAN node access data from that node at local storage speeds, while I get the impression with HP's VSA that it presents as iSCSI and thereby subject to that slow down.
I like the HP VSA because it would allow us more than 3 nodes and it allows the migration to hardware based SAN or even mixed hardware and VM based nodes, but if I am understanding the performance difference with VMware's product I would consider that a greater value and worth the 3-node limit.
My question is with HP LeftHand VSA, since you configure targeting the cluster and not the node, will it not be able to bi-pass the artificial performance ceiling of the virtualized switch and NIC in the communication chain between the local VM and local virtual SAN node?
Using the iSCSI protocol, the Lefthand VSA will add some latency to any disk IO requests processed through it. Unless you're looking for a high performance solution, this shouldn't be noticeable compared to the latency of the storage medium however. If you're looking for some sort of high performance solution, a VSA is probably the wrong way anyhow.
Personal experience is that storage visualization still needs time to mature. There are speed tests on Google which show a significant overhead with storage virtualization on the VMware vSA platform. I have not seen any HP LeftHand VSA performance metrics though.
VMware does provide a while paper which goes over a few metrics, Performance of VSA in VMware vSphere 5. It dipicts response time in the 1000ms range. It looks like VMware only supports a RAID-10 configuration for performance concerns.