vSwitch configuration with 12 uplinks

Solution 1:

Use separate vSwitches with dedicated uplinks for Management, Data, and iSCSI traffic. You can share vMotion and Management in most cases, but if you have enough NICs to separate them, feel free.

You don't need to do trunking in this configuration, just have the physical ports in the correct VLAN and you should be all set. If you have multiple data VLANs that you want your VMs to be on, then you can trunk the data ports and configure the vSwitches appropriately, but it doesn't sound like this is what you need.

Solution 2:

If you've not already spent the money then consider just buying two 10Gbps NICs instead and running everything over those, it'll be a hell of a lot easier to setup and maintain, and I can vouch for it's simplicity and performance.

Solution 3:

MOAR isn't necessarily better :)

Here's a perfectly-resilient 6-NIC ESXi host setup, using:

2 pNICs - VMWare VM traffic.
2 pNICs - VMWare VMotion traffic.
2 pNICs - NFS storage. Could also be iSCSI traffic with MPIO.

You can trunk (I did in this case because the environment is multi-tenant), or you can tag individual ports with their respective VLAN. In the example here, each vSwitch have physical NICs plugged into a different member of a Cisco 3750 switch stack. We're also using diversity across physical NIC cards. So we can lose one HBA, link, switch, etc. and still pass traffic.

If you're still using 1GbE links, I'm not sure much more can be gained by utilizing more physical NICs (e.g. 12!). However, that depends on what you need your VM's to do.

Solution 4:

You should configure at least one dedicated NIC for iSCSI traffic. Best practice suggests that you also dedicate a separate NIC each for Virtual Machine traffic, vMotion and Fault Tolerance. Best practice also suggests that you have two VMkernel ports configured for Management traffic.

As for the pSwitch you shouldn't need to use trunk ports. The pSwitch ports should be configured as access ports with membership in the appropriate VLAN.