vSphere NIC Teaming - Any reason for Active/Standby?

Question

We're using ESXi and I'm wondering about the "NIC Teaming" options for our vSwitches.

On a particular vSwitch with 2 NICs, is there any GOOD reason for configuring a NIC as Active and the other as Standby?. I always use Active/Active scenarios (don't see any reason for Active/Standby because if one NIC fails (and ESXi notices it) it will switch all traffic to the other NIC anyway.

Also, are there any recommendations for the VMkernel port group (or for the vSwitch it resides) as to always use Active/Standby rather than Active/Active?

Thanks! Craconia

Answer 1

The only thing that comes to mind is that in case of a (single) link failure, you don't lose capacity and your application becomes more predictable.

For example link usage can slowly creep up beyond a single link (application growth over months) without anyone fully realizing that. Then during a failure the application is congested and acts up, making it much more difficult to trace the source.

Answer 2

Not familiar with ESXi bonding in particular but I bet it uses Linux bonding. One of the main reasons to use Active/Backup mode is when he NIC connects to a different switch it can be more simple. In general the Active/Standby model is the most straight forward and keeps thie KISS principal. So if you don't need anywhere need the bonded bandwidth some like to just keep it simple.

Answer 3

One basic reason: the switch does not support Active/Active configuration.

Answer 4

Don't mean to be the "well actually" guy, but...

Well actually, the most commmon reason is because each nic might go to a different switch. This is useful especially in ethernet based storage networks where redundancy is required at the host, switch, and filer levels. We are setup in exactly this way

Answer 5

Your network switch must be properly configured to use an active-active nic teaming.. in Cisco switches you need to configure a port-channel. For further information an a clear overview check this blog post: http://blog.scottlowe.org/2006/12/04/esx-server-nic-teaming-and-vlan-trunking/

Answer 6

You might want to try to keep all communication within a single switch in normal operating conditions, so that you do not put any unneccessary load and risk congestion on a link between the switches, instead keeping the traffic within a single switch backplane. This could reduce latency and eliminate a bottleneck, depending on the way your network traffic patterns look.