NIC teaming and aggregate throughput

[zenomorph]

Hi,

I need some help on configuring NIC teaming on our ESX. We have a HP Server with 6 NICs I already configured the vSwitch for VMkernal, VMotion and Service Console.

Now since the ESX will be running various Web Servers what I want to do is aggregate the throughput of the 1GB NICs to give me a total of 4GB throughput. What I've done is created a vSwitch and alllocated the 4 NICs to the switch and created a port called "Prod LAN" and assigned the 4 NIC, also on our Cisco switch I have trunked the 4 ports.

What else do I need to do to ensure I have correctly configured the NICs so that when we assign the NICs to the Web Server VM it can use the aggregated 4GB throughput. Also what I also want to ensure is if any of the 1,2 or 3 NICs fail I still have throughput on the port.

Many thanks.........

[weinstein5]

What you will need to do is select the IP Hash NIC Teaming method to insure that the VMs will be able to transmit across any of the four ports - redundancy by default is built in when you configure a NIC team.

If you find this or any other answer useful please consider awarding points by marking the answer correct or helpful

[zenomorph]

Weinstein,

If I use IP Hash NIC Teaming does that mean I will actually be able to utilize all the NICs and get 4GB throughput on a VM if I bind a "Prod LAN" port to the VM?

Many thanks

[weinstein5]

Kind of - the way IP Hash works is the vmkernel examines every outbound packet and based upon the originating IP and destination IP will choose the a physical NIC - so in your scenario with web servers the outbound traffic will have the high probability going out of any one of the four physical NICs -

If you find this or any other answer useful please consider awarding points by marking the answer correct or helpful

[depping]

there's a great explanation found here: http://kensvirtualreality.wordpress.com/2009/04/05/the-great-vswitch-debate%E2%80%93part-3/

Duncan

VMware Communities User Moderator | VCP | VCDX

-

---

Blogging:

Twitter:

If you find this information useful, please award points for "correct" or "helpful".

[joshp]

As long as the four trunks are connected to one physical switch you must enable 802.3ad link aggregation (Ether-channel) on the Cisco switch. This will then allow an aggregation of the four 1GB NICs to allow up to 4GB throughput. Also, once the Cisco switch is configured for 802.3ad link aggregation on the four trunks you set the vSwitch teaming method to IP Hash.

[zenomorph]

Guys,

Many thanks for your reply.

But just want to ask one more thing, alternatively what if I want o configure as follows;

On the vSwitch allocate 4NICs, and our datacentre has 2 Core Switches so what I might do is team 2 NICs to core switch A and then 2 NICs to core Switch B;

NIC1 and NIC2 ->CoreSwitch A trunk NIC1/2 802.3(eg. Team A)

NIC3 and NIC4 -> CoreSwitch B trunk NIC3/4 802.3 (eg. Team B)

then how can I configure it to use default Team A by default and if the link fails use Team B?

Much apprecaited............

[AndreTheGiant]

then how can I configure it to use default Team A by default and if the link fails use Team B?

You can do using active and standby adapter, instead of all active adapters.

You can change this policy or at vSwitch level, or at port group level.

Andre

**if you found this or any other answer useful please consider allocating points for helpful or correct answers

[weinstein5]

I do not think you can

Sent from my iPhone

On Jun 14, 2009, at 12:24 AM, "zenomorph" <communities-emailer@vmware.com

[joshp]

I agree, you cannot aggregate links from two separate switches to a single vSwitch unless your two switches share their backplane (usually stackable switches).

[depping]

never ever mix channels on a vSwitch. when one of the links of the active channel fails 1 link of the passive channel will be added. ESX isn't aware of these channels and the switch will go nuts because traffic is appearing on places it's not supposed to appear.

Duncan

VMware Communities User Moderator | VCP | VCDX

-

---

Blogging:

Twitter:

If you find this information useful, please award points for "correct" or "helpful".

[zenomorph]

Oh ic.....

Not sure if my questions was clear, actually

NIC1 and NIC2 -&gt;CoreSwitch A trunk NIC1/2 802.3(eg. Port 1 - Team A)

NIC3 and NIC4 -&gt; CoreSwitch B trunk NIC3/4 802.3 (eg. Port 2 - Team B)

what I was thinking was default all VMs use default NIC Port 1 which on our Cisco switch I trunk them, then Port 2 I use as a failover NIC. So in effect what I want to do is run aggregate the throughput of each NIC team to 2GB such that if that Core switch fails I can failover to our other Core switch which also has 2 NICs trunked so I can maintain 2GB throughput.

Or is it such that it can only be done with a single NIC and failover eg.

NIC 1 Core switch A

NIC 2 Core switch B

configure the vSwitch with failover settings between NIC 1 and NIC 2.

Many thanks

[joshp · Accepted solution]

Duncan identified the issue above with what you are proposing. You might be okay with your config if an entire physical switch fails but as identified above, if you lose a single link on one of the aggregates ESX will add a single link from the passive aggregate causing the switch to go crazy.

[zenomorph]

Joshp...

Now I know what you mean, many.

many thanks for your great help guys................