Implementing IBM System Networking 10Gb Ethernet Switches

Chapter 2. IBM System Networking Switch 10Gb Ethernet switch features 77

2.7 High availability

Internet traffic consists of myriad services and applications that use the Internet Protocol (IP)

for data delivery. However, IP is not optimized for all the various applications. High availability

goes beyond IP and makes intelligent switching decisions to provide redundant

network configurations.

2.7.1 Trunking

When using port trunk groups between two switches, as shown in Figure 2-12, you can create

a virtual link between the switches, operating with combined throughput levels that depends

on how many physical ports are included.

Figure 2-12 Trunking

Two trunk types are available:

򐂰 Static trunk groups (portchannel)

򐂰 Dynamic LACP trunk groups

Trunk groups are also useful for connecting a G8264 to third-party devices that support link

aggregation, such as Cisco routers and switches with EtherChannel technology (not ISL

trunking technology) and the Sun Quad Fast Ethernet Adapter. Trunk group technology is

compatible with these devices when they are configured manually.

Trunk traffic is statistically distributed among the ports in a trunk group, based on various

configurable options.

Also, because each trunk group is composed of multiple physical links, the trunk group is

inherently fault tolerant. If one connection between the switches is available, the trunk

remains active, and statistical load balancing is maintained whenever a port in a trunk group

is lost or returned to service.

Configurable trunk hash algorithm

Traffic in a trunk group is statistically distributed among member ports by using a hash

process where various address and attribute bits from each transmitted frame are

recombined to specify the particular trunk port the frame uses.

The switch can be configured to use various hashing options. To achieve the most even

traffic distribution, select options that exhibit a wide range of values for your particular

network. Avoid hashing on information that is not present in the expected traffic, or which

does not vary.