So you have PCI-E FLASH,RAMSAN on your VSA but can't push each LUN to 255 QD (or more?)
2 vcpu with default settings are not going to push that 40gbe nic and storage subsystem to 1 million IOPS!
The problem is of course simple - if you can't push 255 QD your SSD will never get a true workout. If you can't dispatch i/o against more VCPU, you will not be able to handle the i/o - Both EThernet and Storage this is true!
What I would suggest is the ability to tweak both storage and networking for the VSA - esxi -> and hyper-v so you can push hard.
I would suggest that a STOREVIRTUAL VSA 11 with TIER-0 of RAM [backed with flash], and TIER-1 of SSD might be able to reach 1 million IOPS with the ability to add more VCPU, tweak Storage Queue Depth/Ring, and likewise tweak the VMXNET3 subsystem to run 4,5,6,7,8 (or more) vcpu.
So here's the deal! If anyone can get logged into their lefthand (shell) VSA..
try this - or maybe HP lefthand engineers can tell us the whole story.......
[I am asking lefthand engineers, not end users ..! DO NOT TRY THIS AT HOME KIDS! ]
Append the following to kernel boot args ( edit /boot/grub/grub.cfg) and reboot the system.
vmw_pvscsi.cmd_per_lun=255 vmw_pvscsi.ring_pages=32
You can verify the changed queue depth by doing a cat on these module args i.e.
$ cat /sys/module/vmw_pvscsi/parameters/cmd_per_lun
$ cat /sys/module/vmw_pvscsi/parameters/ring_pages
(link to older source: http://permalink.gmane.org/gmane.comp.emulators.qemu/142160)
#define PVSCSI_DEFAULT_NUM_PAGES_PER_RING 8
#define PVSCSI_DEFAULT_NUM_PAGES_MSG_RING 1
#define PVSCSI_DEFAULT_QUEUE_DEPTH 64
#define SGL_SIZE PAGE_SIZE
