Configuration of vNIC / Loadbalancing / Failover / QOS
Below is a typical configuration of vSphere ESX vNIC assignment correlating to the UCS Fabric Interconnect.
|Management Port Group||vMotion Port Group||Fault Tolerance Port Group||Production Virtual Machine Port Group||Storage HBA’s|
|UCS Fabric A 10GB||vNic 0 (Active)||vNic 2 (Standby)||vNic 4 (Active)||vNic 6 (Active)||HBA 1 (Active)|
|UCS Fabric B 10GB||vNic 1 (Standby)||vNic 3 (Active)||vNic 5 (Standby)||vNic 7 (Active)||HBA 2 (Active)|
|UCS QOS Priority||Bronze||Silver||Gold||Best Effort||FC|
|UCS QOS Weight %||0 %||11 %||11 %||33%||34 %|
|Min / Maxbandwidth using both fabrics||Min = 1GBMax = 1GB||Min = 1GBMax = 10 GBMax is limited to QOS Priority||Min = 1GBMax = 10 GBMax is limited to QOS Priority||Min = 6GBMax = 10 GBMax is limited to QOS Priority||Min = 8GBMax = 10GBMax is limited to QOS Priority|
|VMware Load Balancing||Route based on originating virtual port Id||N/A||N/A||dVS = Route based on physical NIC loadvSS = Route based on originating virtual port Id
|Active Active(round Robin)|
|VMware Failover Policy||Active/Standby||Active/Standby||Active/Standby||Active / Active||N/A|
1) Note, that route based on IP hash is not supported by UCS when used with either a VMware Standard vSwitch or Distributed vSwitch. You can use route based on IP hash when used with a nexus 1000v.
2) With the above configuration the production vNics are both active and separated between two fabrics. The uplinks from the fabrics should then be configured as a Multi Chassis Ether-channel to the core switches.
3) By applying UCS QOS policies, there is guaranteed minimum bandwidth assigned per vNic, and also has the ability to burst higher if is ever needed, but will not affect other vNic’s that have a higher priority.
4) If a dVS (VMware distributed virtual) switch is used, then the “route based on physical NIC load “ load balancing algorithm should be used to load balance active/active uplinks to the fabric interconnects.