What is the best Network setup with 4 NICs (2x 10GB and 2x 1GB)

[bruDan]

I have 4 hosts running ESXi 4 Enterprise licence with 2x 10GB and 2x 1GB NICs, what is the best network setup to use in this setup? where should FT logging go? I cant add more NICs as this is on HP Blades.

This is my current setup:

Thanks

Dan

[a_p_]

Depending on how many VM's you will have, how much network traffic you expect, how many VM's are going to be FT enabled... I'd suggest the following:

A.) vSwitch0

Physical NIC's: vmnic2 + vmnic3 (10GB/s)

A.1) VMkernel Port Group (Management Network)

vmnic2 = active

vmnic3 = standby

A.2) VMkernel Port Group (Vmotion)

vmnic2 = active

vmnic3 = standby

A.3) Virtual Machine Network(s) Portgroup

vmnic2 = standby

vmnic3 = active

B.) vSwitch1

Physical NIC's: vmnic0 + vmnic1 (1GB/s)

B.1) VMkernel Port Group (Fault Tolerance)

vmnic0 = active

vmnic1 = active

To separate the traffic, use different VLANS for the port groups.

This suggestion may not be best practice, however with 4 NIC's that's what I would probably do.

IMHO on vSwitch0 you could also leave both vmnic's active. 10GB/s should be able to handle this.

André

[bruDan]

Hi André,

I dont think we will be having any more than 4 VM's FT enabled, and i think around 40 VM's possibly less.

The main VM's will be running SQL Server.

Would i need to setup trunks on the blade switches for the active/active teamed vSwitches?

The 2x 10GB nics on each host are connected to one blade switch and the 2x 1GB nics connected to another (this how the blade enclosure maps them and cannot be changed) would it be possible/feasible to configure it so that if one blade switch where to fail the VM's would continue to run over the slower 1GB links? or am i asking too much?

thanks

Dan

[a_p_ · Accepted solution]

You are talking about the c3000 blade enclosure, aren't you?

For a disaster tolerant solution, you will probably go with:

A.) vSwitch0

Physical NIC's: vmnic0 (1GB/s) + vmnic2 (10GB/s)

A.1) VMkernel Port Group (Management Network)

vmnic0 = standby

vmnic2 = active

A.2) VMkernel Port Group (Vmotion)

vmnic0 = active

vmnic2 = standby

A.3) Virtual Machine Network(s)

vmnic0 = standby

vmnic2 = active

B.) vSwitch1

Physical NIC's: vmnic0 (1GB/s) + vmnic3 (10GB/s)

B.1) VMkernel Port Group (Fault Tolerance)

vmnic1 = standby

vmnic3 = active

VMware - by default - uses round robin to assign a VM to a vmnic.

So you don't setup port aggregating on the switches.

You should, however, separate traffic using VLAN's if possible, therefore you need to configure the ports as trunk ports.

André

[thehyperadvisor]

I would also do what Andre suggested based on the current situation. I typically would have the same switch modules for like port type and speed when teaming in general but it will work. I would also only trunk the uplink ports on the switch modules and not the vmnics or ports on the midplane of the enclosure. And I also enable "Beacon Probing" in the team with blades so that the esx host knows when the switch uplink ports are not connected or down.

hope this helps - thehyperadvisor.com

If you found this or other information useful, please consider awarding points for "Correct" or "Helpful".

[bruDan]

Hi,

Sorry for the delay, yes its the c3000 blade enclosure.

Thanks for the suggestions.

When you say trunk the uplinks ports on the modules do you mean the links between the blade switches and the main network as apposed to the links to the vSwitches?

[thehyperadvisor]

That is correct.

hope this helps - thehyperadvisor.com

If you found this or other information useful, please consider awarding points for "Correct" or "Helpful".