I want to convert an iSCSI vsphere setup to use NFS instead (makes pacemaker cluster simpler, since all the failover code needs to do is manipulate the virtual IP and import/export the ZFS pool the NFS share lives on.) I am currently using a 10gbe link from the two vsphere hosts to the iSCSI storage server, with a 1gbe link as backup. I have the preferred data path for the vmfs datastore set as the 10gbe link. Works fine. Now that vsphere has NFS 4.1, I though of using the new multiple path capability, but as far as I can tell, it's a simple round-robin algorithm, so I won't get remotely 10gb out of this. The only trick I can think of is to define several virtual IP addresses aliases on the NFS server's 10gbe nic. Is there a better way (this seems like an awful kludge). For now, I'm inclined to stick with a simple mode 1 bond, but that only detects simple link failure. I've also seen references to "Route based on physical NIC load", which sounds like mode 5 (balance-tlb bonding?), although that requires a distributed vswitch on the vsphere hosts (might still be worth looking into at some point, though...) Any thoughts/tips appreciated