Etherchannel ESX 4.1 and Equallogic

This is a user-to-user forum. If you want an answer from Dell as such, I suggest you call Equallogic support and open a case with them and ask the question.

As a non-Dell answer, my wording may not be 100%, but the general idea is that iSCSI uses MPIO. From a host, each NIC/VMkernel generates an iSCSI session to the group. The group then forwards that session to a physical port on the member. So, with 2 NICs/VMkernels on 2 separate IP addresses, that yields 2 iSCSI sessions where each session generally ends up being forwarded to a different physical port on the member. The MPIO part then splits the SCSI commands (inside of TCP traffic) and sends them to the array. This is how you get a 2Gbit/s 'connection'. If you start teaming NICs, the iSCSI connection becomes a single connection, which gets forwarded to a single physical port on the member (which is a 1Gbit/s port (I'm assuming you don't have a 6010 or 6510)), and therefor you just reduced your bandwidth to 1Gbit/s.

Good morning,

Equallogic arrays do not support Etherchannel / bonding / trunking, etc...    So setting them up provides little advantage. A 2 port channel group, will only create one iSCSI session to the array.   Break that into two MPIO connections and you double your bandwidth and retain the same level of redundancy. 

With ESX v4.x you can enable MPIO which will work hand-in-hand with the load balancing algorithm, One problem with trunking/bonding/etc is the spanning tree overhead as you try to scale connections.   In a 16 member group, with 4x GbE per array, that would be 64 ports that would need to be trunked.   PLUS the trunking from the servers.   Very few if any switches would easily handle that load.   The connection balancing routine we (CLB),  scales w/o having to change the host or switch.   Add another array and those additional connections are used automatically.  No changes needed at the host layer. 

Regards,

-don