Types of vifs

Kinds of vifs on a filer

There are two kinds of vifs:

Single-mode
Multi-mode

Single-mode vif operation

In a single-mode vif, only one of the interfaces in the vif is active. The other interfaces are on standby, ready to take over if the active interface fails. Failure means that the link status of the interface is down, which signals that the interface has lost connection with the switch.

There can be more than one interface on standby in a single-mode vif. If an active interface fails, the filer randomly picks one of the standby interfaces to be the next active link.

All interfaces in a single-mode vif share a common Media Access Control (MAC) address.

Note
A single-mode vif does not require any additional switch configuration or a switch that supports link aggregation.

Example

In the following figure, e0 and e1 are part of the SingleTrunk1 single-mode vif. The active interface, e0, fails. The standby e1 interface takes over and maintains the connection to the switch.

Multi-mode vif operation

In a multi-mode vif, all interfaces in the vif are active and share a common MAC address. This logical aggregation of interfaces provides higher throughput than a single-mode vif. Multi-mode vifs can recover from a failure of up to (n-1) interfaces, where n is the total number of interfaces that form the vif.

Note
The multi-mode vif implementation in Data ONTAP is in compliance with IEEE 802.3ad (static).

A multi-mode vif requires a switch that supports link aggregation over multiple switch ports. The switch is configured so that all ports to which links of a vif are connected are part of a single logical port. For information about configuring the switch, see your switch vendor's documentation.

Some switches might not support link aggregation of ports configured for jumbo frames. For more information, see your switch vendor's documentation.

Several load-balancing options are available to distribute traffic among the interfaces of a multi-mode vif. The load-balancing schemes are discussed in detail in Load balancing in multi-mode vifs.

Example

In the following figure, e0, e1, e2, and e3 are part of the MultiTrunk1 multi-mode vif. All four interfaces in the MultiTrunk1 multi-mode vif are active.

If any three of the interfaces fail, either one by one or simultaneously, the filer still stays connected to the network.

Load balancing in multi-mode vifs

To ensure that all interfaces of a multi-mode vif are equally utilized for outgoing traffic, the following load-balancing methods are available:

IP-address based
MAC-address based
Round robin

The load-balancing method to use for a multi-mode vif can be specified only when the vif is created. If no method is specified, the IP-address based load balancing method is used.

IP-address and MAC-address based

In both of these methods, the last byte of the source and destination address (IP address and MAC address) is used to determine the interface to use for the outgoing frame. The following formula is used:

((source_address XOR destination_address) % number_of_links)

If the result of this formula maps to an interface that is not in the UP link-state, the next active interface is used.

For example, a vif consisting of eight physical interfaces is created with the IP-address based load-balancing method. It is configured with IP address 10.0.0.10. Based on the above formula, an IP frame going through this vif to the destination IP address 172.26.15.224 will use interface #2, provided that this interface is in the UP link-state.

Note
You should not select the MAC-address based load-balancing method when creating vifs on a filer that connects directly to a router. In such a setup, for every outgoing IP frame, the destination MAC address will be the MAC address of the router. As a result, only one interface of the vif will be used.

Round robin:

Unlike the IP-address and MAC-address load-balancing methods, this method provides true load balancing. This method may cause out-of-order packet delivery and retransmissions due to overruns.

This method of load balancing is recommended for clients connected in a back-to-back configuration with a filer.