A Cisco engineer is analyzing a customer environment where dozens of VMs are running on...

A Cisco engineer is analyzing a customer environment where dozens of VMs are running on a physical server and UCS chassis are cascaded The setup uses two ISLs in a port channel to the SAN switch (per fabric) The oversubscription from VM to physical SAN port is often 20 (VMs/host) x 72 (9 UCS chassis cascaded = 72 hosts): 2 (ISLs from fabric interconnect to SAN switch port on the NPV core (NPIV enabled switch) This results to 20 x 72 2 per single ISL = 720 VMs per ISL

20 (VMs per host) x 8 (hosts per chassis) x 9 (chassis per port channel to SAN switch) / 2 (ISLs per port channel) = 20 x 8 x 9 / 2 = 720 VMs per SAN F-port What is the best way to resolve possible IOPS bottleneck?