Best design? 6 pNICs and 4 network segments

[FrostyatCBM]

I have 3 x ESXi 4.1 hosts, each with 6 physical NICs:

• 4 on the motherboard

• 2 on an extra card

I need to connect them to 4 physical network segments:

• LAN

• DMZ

• MANAGEMENT (vCenter/vmKernel traffic)

• VMOTION

After thinking about it some, and assuming that I don't just buy some additional 4-port NIC cards ... I was thinking of setting it up like this:

• 4-port motherboard NICs:

• pNIC0 -- LAN

• pNIC1 -- DMZ

• pNIC2 -- MANAGEMENT

• pNIC3 -- VMOTION

• 2-port extra card NICs:

• pNIC4 -- LAN

• pNIC5 -- MANAGEMENT

That leaves me with the following scenarios:

• if the 4-port NICs simultaneously fail (motherboard fault) I can stll control the servers (pNIC5) and the bulk of my data traffic will continue (LAN via pNIC4)

• if the 2-port card fails, pNICs 0-3 keep everything working

• if individual NICs fail, then the things I can most "do without" are DMZ and VMOTION in the short term

Does that sound reasonable?

I'm choosing physical separation of my networks as a design goal, so don't want to go down the path of VLAN tagging.

Should I just bite the bullet and buy 3 more 4-port cards so that I have full redundancy?

[idle-jam]

That would be fine, as if you have VMware HA it would be able to failover to the surviving host. Again do note that you can acquire more NICs but you have the make sure the switch end are redundant as well.

Another way is if you switch supports link bond/aggregation, you could combined the 6 into a 1 big pipe and have VLAN to separate your network.

iDLE-jAM | VCP 2, VCP 3 & VCP 4

If you found this or any other answer useful please consider the use of the Helpful or correct buttons to award points.

[FrostyatCBM]

Yeah, good point on the issue of switch redundancy. At the moment I'm just using a single physical switch per network segment, but if any of them ever fail I have spares on hand which I can swap in, so the downtime will be fairly minimal. They're good quality Cisco switches too, not cheapies, so I hope their MTBF is big!

[idle-jam]

Great to hear that you have a cisco switch, perhaps you can look into Cisco's link bonding - etherchannel and bond all network port instead of purchasing additional ports. The money saved could be use in buying components to beef up the virtualization host performance - say Memory?

iDLE-jAM | VCP 2, VCP 3 & VCP 4

If you found this or any other answer useful please consider the use of the Helpful or correct buttons to award points.