RFC on my iSCSI MPIO network design

[mgbowman]

Please see the attached diagram.

I am currently redesigning my ESXi infrastructure and have been using NFS (in a VLAN) over LAGs and the network performance between my ESXi servers and NAS is terrible. Since I am using a 2-port gigabit LAG for everything, I belive there's just too much on the wire. What I would like to do is switch to dedicated iSCSI MPIO for increased throughput. One thing I thought about was skipping the switches on the iSCSI paths altogether and just directly connecting my ESXi servers to my NAS across NICs. This should give me a "pseudo-dedicated" 2Gbps link to my storage from each ESXi server. The clients would be served through the 1Gbps LAG going to the stacked switches.

Any comments are greatly appreciated.

Thanks in advance!

--mgbowman

[rickardnobel]

mgbowman wrote:

One thing I thought about was skipping the switches on the iSCSI paths altogether and just directly connecting my ESXi servers to my NAS across NICs.

Going without connecting through the physical switches might be fine, but also make this solution very hard to expand in the future.

You should be able to do something very similar - with dedicating physical ports just for iSCSI, but still pass through the physical switches. If they are modern and capable the delay introduced by switches is typically extremely low.

[mgbowman]

I did just that - went through the switches. Made more sense in the end.

Thanks!

--mgbowman