IBM Nortel Networks L2/3, Руководство пользователя

Страница: 77 / 190

Chapter 7. Nortel Networks L2/3 GbESM configuration and network integration
63
Figure 7-1 What trunk failover can protect against
7.3.2 Introduction to NIC Teaming
NIC Teaming is a function that is provided by Broadcom, the manufacturer of the NIC chips
used on the Blade Servers, in their software. Broadcom provides the Broadcom Advanced
Services Protocol (BASP) which includes NIC teaming, as well as the Broadcom Advanced
Control Suite (BACS) which is a Windows application which helps configure NIC teaming.
NIC teaming allows two or more physical NICs to be treated as a single logical network object
in Windows or a single /dev/eth file in Linux. The single object or file can then be assigned
network properties such as an IP address in the same way as any other NIC.
The BACS application allows several types of teams to be created. For HA designs, the
Smart Load Balancing (SLB) team is used. Layer 2 designs can have both of the adapters (on
an HS20 blade) as active members of the team; for Layer 3 designs, an active or standby
team is used with one adapter as an active member of the team and the second adapter as a
standby member of the team.
NIC teaming is intended to provide both additional capacity (bandwidth) as well as High
Availability. The team will detect loss of signal on any of its member NICs and continue to
send traffic through other active members, or activate standby members if necessary. In the
IBM
Eserver
BladeCenter, NIC teaming will detect the failure of a NIC chip on the server
blade, the loss of connection to a switch module via the midplane, and the failure of a switch
module (including intentional removal or power-off). Of these, intentional removal or power-off
of a switch module is by far the most common.
The BASP drivers also provide support for 802.1q tagging of the server NIC. This allows
support for multiple VLANs on a single physical NIC or on a group of teamed NICs. When this
capability is used, each VLAN has its own network object (windows) or /dev/eth file (Linux).
Thus, each VLAN can be assigned its own IP address. This can be useful to isolate different
categories of traffic from each other or to provide different Quality of Service (QoS)
configurations for different types of traffic whose target is the same server. A sample
If failure anywhere on the link toward the
upstream switch, the NIC on Blade server
does not know about the failure and may
continue to send traffic toward the top switch,
which will discard the traffic. - The Trunk
Failover feature addresses this issue
.
If the switch fails in such a way that the link
toward the Blade server goes down, or NIC
fails, Blade server can sense this and
redirect traffic out the other NIC toward the
bottom switch. NIC Teaming can take care of
this without the need for trunk failover.
1 2 3 4 5 6 1 2 3 4 5 6
GbESM
GbESM
X
X
X
VLAN X
Teamed
Active/Standby
NIC1 NIC2
Logical NIC Interface
X
X
Blade server 1
BladeCenter
Topology 1 - Trunk Failover