Load balancing distributes the network load across multiple available adapters. Load balancing policies are configured as a part of the NIC Teaming and failover options on virtual switches, virtual machine PortGroups, and VMkernel interfaces.

The following screenshot illustrates the Edit Settings dialog for configuring the Teaming and failover options on the virtual machine PortGroup on a standard virtual switch:

The following load balancing policies can be applied to virtual switches or virtual PortGroups:

• Route based on originating virtual port: This is the default load balancing policy. When it is being used, the load is balanced based on the number of physical NICs and the number of virtual switch ports in use. Virtual port connections are distributed across the physical NICs available to the virtual switch. A virtual machine connected to the virtual port will always use the same physical NIC, unless the NIC becomes unavailable. This is usually the best load balancing method to use.
• Route based on IP hash: This load balancing policy uses a hashing algorithm that determines the physical path based on the source and destination IP addresses of the virtual machine traffic. A virtual machine's network traffic can use multiple available NICs. This policy is used when using either EtherChannel or the LACP link aggregation.
• Route based on source MAC hash: This load balancing policy is similar to the Route based on originating virtual port policy, except that the physical NIC used for virtual machine traffic is based on the virtual network adapter's MAC address and not the virtual port connection.
• Use explicit failover order: This policy is not really a load balancing policy, because network traffic always uses the physical NIC uplink that is configured as the highest ordered active physical uplink available.
• Route based on physical NIC load: This is an additional load balancing option offered by vDSwitches that is not available to vSwitches. It is the most efficient because it distributes the load across active uplinks, based on the actual workload of the physical NICs.

Redundancy in the virtual network is provided by configuring the Failover order. These configurations define the physical uplinks that are actively used to pass network traffic, and those that are available stand in the event of an active uplink failing.

The available adapters are as follows:

• Active adapters: These are adapters that are available for use by the virtual switch, a virtual machine network PortGroup, or a VMkernel interface.
• Standby adapters: These are adapters that only become active in the event that an active adapter becomes unavailable.
• Unused adapters: These adapters are unused. They will never be used by the virtual switch, a virtual machine PortGroup, or a VMkernel interface.

The following screenshot shows the adapters that are configured in Active and Standby. If the Active Adapter (vmnic0) fails, the Standby Adapter (vmnic1) will become active:

The following diagram shows an example of an active/standby network configuration that's commonly used in small environments to provide connectivity and redundancy for the host management and vMotion networks:

Network Failover Detection and Failback are settings that control how a network failure is detected, and what happens when an active adapter is returned to service after a failure.

How network failure is detected is configured by using the Network Failover Detection option. Two failure detection options are available: Link Status Only and Beacon Probing. The Link Status Only option uses the link state (up or down) of the physical NIC to determine whether the uplink is available. The Beacon Probing option detects network failures by sending and receiving beacon probes out to all physical uplinks on the virtual switch, and it can detect link state and switch failures. At least three active uplinks are needed for beacon probing to work effectively. The VMware Knowledge Base article located at http://kb.vmware.com/kb/1005577 provides more information on how beacon probing works with virtual switches.

The Failback setting defines whether an active adapter is returned to service if the adapter becomes available after a failure, based on the value set for Network Failover Detection. If a physical switch fails and Failback is enabled, with Link Status Only being used for Failover Detection, the adapter may become active and will be returned to service before the physical switch is available to pass traffic.