BladeSystem Virtual Connect
cancel
Showing results for 
Search instead for 
Did you mean: 

VMware 5 port profile settings and monitoring load

 
chocky005
Occasional Visitor

VMware 5 port profile settings and monitoring load

I Have 2 HP 7000 Blade chassis linked with ProLiant BL460c G7 blades running ESXi5 connected to the network with  HP VC Flex Fabric 10gb/24-Port Modules. 

 

The Blades have the following Port porfile config.

6 nic’s running on 6gb

2 HBA running on 4gb

 

Nic 1 ESX Console and Nexus 1000v – bay1 running at 1gb

Nic 2 ESX Console and Nexus 1000v – bay2 running at 1gb

Nic 3 Main VM Network – bay1 running at 3.5gb

Nic 4 Main VM network – bay2 running at 3.5gb

Nic 5 Vmotion – bay1 running at 1.5gb

Nic 6 Vmotion – bay2 running at 1.5gb

 

HBA 1 Fabric_A  - Bay1 running at 4gb

HBA 2 Fabric_B – Bay2 running at 4gb

 

Questions

1. Is this the Best or Recommened setup for the ESXi Hosts in the HP chassis?

2. How can I monitor and see if one interface is under more load than the others?

 

1 REPLY
Casper42
Trusted Contributor

Re: VMware 5 port profile settings and monitoring load

With VC Firmware 4.0 you get 2 things that may help you.

1) You can monitor the performance of individual FlexNICs in the profile using a 3rd party SNMP based management tool. The new MIBs have been on the website for 6+ months at least.

2) You can now allow individual FlexNICs to burst. So your 1.5Gb for vMotion, you can now set a Max speed of 5Gb and this will allow the vMotion NICs to burst to 5Gb, get the vMotion done quickly and drop back to idle. Then that 1.5 can be used by other FlexNICs like for the VMs when vMotion is idle.

Aside from that, I've found the easiest way to monitor interface utilization is actually from the VMware side. It gathers a lot of these metrics already and you can see if a particular NIC is getting close to its max you have set in the profile.

You only need to monitor TRANSMIT Bandwidth though because VC does not limit Receive bandwidth. By the time the packet reaches the NIC, which is where the bandwidth throttling happens, it doesn't make a lot of sense to drop it and make the sender retransmit.

All that aside, I don't see anything glaringly wrong with your config. We have a ESXi 5 on FlexFabric Guide and your numbers are pretty close to the ones in there.