Does ESXi Link Aggregation require configuration on the physical switch?

[bstafford]

Hi,

I have two physical servers that I'm setting up with ESXi. Each server has eight physcial NICs. My plan is, on each server to use:
- 2 for management - connected to vSwitch0
- 2 for VMotion - connected to vSwitch1
- 2 for iSCSI - connected to vSwitch2 (I'll be using a SAN)
- 2 for VM network access - connected to vSwitch3

For every pair, one cable would be plugged into switch1 and the second would be plugged into switch2.

The switches are HP A5120 that are stacked into one big switch with IRF. i.e. I have one 'big' switch that, if one of the two physical switches dies or gets powered off, would cause the 'big' switch to lose half its ports. This way, all four networks should keep running even if one switch 'dies'.

All the information I have found online says that ESXi can have two physcial NICs attached to a single vSwitch and that it can then do clever stuff for redundancy (in case a physical NIC dies) and load balanacing.

However, while all the guides tell me how to configure it on ESXi, I'm still not sure if I should aggregate the two cables at on the physical switches.

So, my questions are:

1) In order to get NIC redundancy and/or load balancing working in ESXi, do I have to aggregate the two physical ports on the switch(s) that connect to the the two physical ports that connect to the vSwitch?

2) With NIC teaming, I see, for the management network, how having one active adapter and one standby adapter works fine, but, for the iSCSI network, wouldn't I want both adapters to be active to double the bandwidth?

Any insight would be much appreciated.

Thanks.

[Sreejesh_D]

1) In order to get NIC redundancy and/or load balancing working in ESXi, do I have to aggregate the two physical ports on the switch(s) that connect to the the two physical ports that connect to the vSwitch?

>>> for redundancy its not required to create a etherchannel at the physical switch. because if one port/path fails the traffic will flow through the other one.

But for load balancing its a good idea to aggregate physical ports. Because by doing this we will get aggregate throughput of both ports.

2) With NIC teaming, I see, for the management network, how having one active adapter and one standby adapter works fine, but, for the iSCSI network, wouldn't I want both adapters to be active to double the bandwidth?

>> its vary based on the storage vendor and storage. its good to go through the vendor best practice guide for iSCSI settings in VMware.

Like this Dell document.

http://www.dell.com/downloads/global/solutions/vmware_iscsi.pdf

[Sreejesh_D]

1) In order to get NIC redundancy and/or load balancing working in ESXi, do I have to aggregate the two physical ports on the switch(s) that connect to the the two physical ports that connect to the vSwitch?

>>> for redundancy its not required to create a etherchannel at the physical switch. because if one port/path fails the traffic will flow through the other one.

But for load balancing its a good idea to aggregate physical ports. Because by doing this we will get aggregate throughput of both ports.

2) With NIC teaming, I see, for the management network, how having one active adapter and one standby adapter works fine, but, for the iSCSI network, wouldn't I want both adapters to be active to double the bandwidth?

>> its vary based on the storage vendor and storage. its good to go through the vendor best practice guide for iSCSI settings in VMware.

Like this Dell document.

http://www.dell.com/downloads/global/solutions/vmware_iscsi.pdf

[weinstein5 · Accepted solution]

1) In order to get NIC redundancy and/or load balancing working in ESXi, do I have to aggregate the two physical ports on the switch(s) that connect to the the two physical ports that connect to the vSwitch?

No you do not need to configure etherchannel/LACP but it will depend on the type of Load Balancing you select on the ESXi host - if you use default you will not need configure Etehrchannel/LACP since the traffice for each virtual network interface will only com ove a single physical port but you will still the reiciliency in case if one of the physical nics fails the traffic will then go out the remaining port

If you select the IP Hash load balancing you will then need to implement Etherchannel/LACP on your physical switch since traffice can come form any physical port in the team.

2) With NIC teaming, I see, for the management network, how having one active adapter and one standby adapter works fine, but, for the iSCSI network, wouldn't I want both adapters to be active to double the bandwidth?

Well onece again it is going to depend on the Load Balancing methodology you select but you will never be able to bond the physical ports to get the full 2 GB =

I have also moved this to a more appropriate forum -

[bstafford]

Hi weinstein5 and yezdi,

Thank you for such swift responses and for clearing up that question. I guess I now have to study up on load balancing options 🙂