Trunks

A trunk is a logical grouping of interfaces on the BIG-IP^® system. When you create a trunk, this logical group of interfaces functions as a single interface. The BIG-IP system uses a trunk to distribute traffic across multiple links, in a process known as link aggregation. With link aggregation, a trunk increases the bandwidth of a link by adding the bandwidth of multiple links together. For example, four fast Ethernet (100 Mbps) links, if aggregated, create a single 400 Mbps link.

With one trunk, you can aggregate a maximum of eight links. For optimal performance, you should aggregate links in powers of two. Thus, you ideally aggregate two, four, or eight links.

The purpose of a trunk is two-fold:

• To increase bandwidth without upgrading hardware
• To provide link failover if a member link becomes unavailable

You can use trunks to transmit traffic from a BIG-IP system to another vendor switch. Two systems that use trunks to exchange frames are known as peer systems.

In a typical configuration where trunks are configured, the member links of the trunk are connected through Ethernet cables to corresponding links on a peer system.

This figure shows an example of a typical trunk configuration with two peers and three member links on each peer:

Example of a trunk configured for two switches

A primary goal of the trunks feature is to ensure that frames exchanged between peer systems are never sent out of order or duplicated on the receiving end. The BIG-IP^® system is able to maintain frame order by using the source and destination addresses in each frame to calculate a hash value, and then transmitting all frames with that hash value on the same member link.

The BIG-IP system automatically assigns a unique MAC address to a trunk. However, by default, the MAC address that the system uses as the source and destination address for frames that the system transmits and receives (respectively), is the MAC address of the lowest-numbered interface of the trunk.

The BIG-IP system also uses the lowest-numbered interface of a trunk as a reference link. The BIG-IP system uses the reference link to take certain aggregation actions, such as implementing the automatic link selection policy. For frames coming into the reference link, the BIG-IP system load balances the frames across all member links that the BIG-IP system knows to be available. For frames going from any link in the trunk to a destination host, the BIG-IP system treats those frames as if they came from the reference link.

Finally, the BIG-IP system uses the MAC address of an individual member link as the source address for any LACP control frames.

A key aspect of trunks is Link Aggregation Control Protocol, or LACP. Defined by IEEE standard 802.3ad, LACP is a protocol that detects error conditions on member links and redistributes traffic to other member links, thus preventing any loss of traffic on the failed link. On a BIG-IP^® system, LACP is an optional feature that you can configure.

You can also customize LACP behavior. For example, you can specify the way that LACP communicates its control messages from the BIG-IP system to a peer system. You can also specify the rate at which the peer system sends LACP packets to the BIG-IP system. If you want to affect the way that the BIG-IP system chooses links for link aggregation, you can specify a link control policy.

Using the Interfaces setting, you specify the interfaces that you want the BIG-IP^® system to use as member links for the trunk. Once you have created the trunk, the BIG-IP system uses these interfaces to perform link aggregation.

The BIG-IP system uses the lowest-numbered interface as the reference link. The system uses the reference link to negotiate links for aggregation.

After creating the trunk, you assign the trunk to one or more VLANs, using the same VLAN screen that you normally use to assign an individual interface to a VLAN.

The Ether Type property appears in the BIG-IP^® Configuration utility only when the system includes ePVA hardware support. An ether type is a two-octet field in an Ethernet frame, used to indicate the protocol encapsulated in the payload. The BIG-IP system uses the value of this property when an interface or trunk is associated with a IEEE 802.1QinQ (double tagged) VLAN. By default, the system sets this value to 0x8100.

As an option, you can enable LACP on a trunk. Containing a service called lacpd, LACP is an IEEE-defined protocol that exchanges control packets over member links. The purpose of LACP is to detect link error conditions such as faulty MAC devices and link loopbacks. If LACP detects an error on a member link, the BIG-IP^® system removes the member link from the link aggregation and redistributes the traffic for that link to the remaining links of the trunk. In this way, no traffic destined for the removed link is lost. LACP then continues to monitor the member links to ensure that aggregation of those links remains valid.

By default, the LACP feature is disabled, to ensure backward compatibility with previous versions of the BIG-IP system. If you create a trunk and do not enable the LACP feature, the BIG-IP system does not detect link error conditions, and therefore cannot remove the member link from link aggregation. The result is that the system cannot redistribute the traffic destined for that link to the remaining links in the trunk, thereby causing traffic on the failed member link to be lost.

The LACP Mode setting appears on the Trunks screen only when you select the LACP setting. You use the LACP Mode setting to specify the method that LACP uses to send control packets to the peer system. The two possible modes are:

Active mode

You specify Active mode if you want the system to periodically send control packets, regardless of whether the peer system has issued a request. This is the default setting.

Passive mode

You specify Passive mode if you want the system to send control packets only when the peer system issues a request, that is, when the LACP mode of the peer system is set to Active.

If you set only one of the peer systems to Active mode, the BIG-IP^® system uses Active mode for both systems. Also, whenever you change the LACP mode on a trunk, LACP renegotiates the links that it uses for aggregation on that trunk.

The LACP Timeout setting appears on the Trunks screen only when you select the LACP setting. You use the LACP Timeout setting to indicate to the BIG-IP^® system the interval in seconds at which the peer system should send control packets. The timeout value applies only when the LACP mode is set to Active on at least one of the switch systems. If both systems are set to Passive mode, LACP does not send control packets.

If LACP sends three consecutive control packets without receiving a response from the peer system, LACP removes that member link from link aggregation. The two possible timeout values are:

Short

When you set the timeout value to Short, the peer system sends LACP control packets once every second. If this value is set to Short and LACP receives no peer response after sending three consecutive packets, LACP removes the link from aggregation in three seconds.

Long

When you set the timeout value to Long, the peer system sends LACP control packets once every 30 seconds. A timeout value of Long is the default setting. If set to Long and LACP receives no peer response after sending three consecutive packets, LACP removes the link from aggregation in ninety seconds.

Whenever you change the LACP timeout value on a trunk, LACP renegotiates the links that it uses for aggregation on that trunk.

In order for the BIG-IP^® system to aggregate links, the media speed and duplex mode of each link must be the same on both peer systems. Because media properties can change dynamically, the BIG-IP system monitors these properties regularly, and if it finds that the media properties of a link are mismatched on the peer systems, the BIG-IP system must determine which links are eligible for aggregation.

The way the system determines eligible links depends on a link selection policy that you choose for the trunk. When you create a trunk, you can choose one of two possible policy settings: Auto and Maximum Bandwidth.

When frames are transmitted on a trunk, they are distributed across the working member links. The distribution function ensures that the frames belonging to a particular conversation are neither mis-ordered nor duplicated at the receiving end.

The BIG-IP^® system distributes frames by calculating a hash value based on the source and destination addresses (or the destination address only) carried in the frame, and associating the hash value with a link. All frames with a particular hash value are transmitted on the same link, thereby maintaining frame order. Thus, the system uses the resulting hash to determine which interface to use for forwarding traffic.

The Frame Distribution Hash setting specifies the basis for the hash that the system uses as the frame distribution algorithm.

The default value is Source/Destination IP address.

Possible values for this setting are:

Source/Destination MAC address

This value specifies that the system bases the hash on the combined MAC addresses of the source and the destination.

Destination MAC address

This value specifies that the system bases the hash on the MAC address of the destination.

Source/Destination IP address

This value specifies that the system bases the hash on the combined IP addresses of the source and the destination.