Brocade Network OS Administrator’s Guide v4.1.1 - Manual (Page 675)

NIC teaming with vLAG
NIC teaming permits link aggregation between server and switch. It can be one of two types: active/
passive model or active/active model. For the active/passive model, you may not need to configure a
LAG on the switch side, as unique MAC addresses will be seen on only one link.
For the active/active model, the same MAC address may appear on both the links terminating on a
switch (or pair of switches). In such a case, you must configure a LAG on the switch side.
Selecting the MTU
Always set the switch MTU to the maximum host MTU plus 100 bytes. This method is recommended
because the definition of MTU sometimes varies among different vendors. If the switch MTU is set to
the same as the connected host MTU, packets could be dropped.
Avoiding oversubscription
Under certain congestion conditions, you may observe incrementing packet drops representing "tail-
drops" in the output of the show qos rcv-queue interface tengigabitethernet command, as shown
underlined in the following example:
switch# show qos rcv-queue interface tengigabitethernet 5/0/1
Interface TenGigabitEthernet TenGigabitEthernet 5/0/1
In-use 0 bytes, Total buffer 144144 bytes
0 packets dropped
In-use Max
CoS Bytes Bytes
-------------------------
0 0 18018
1 0 18018
2 0 18018
3 0 18018
4 0 18018
5 0 18018
6 0 18018
7 0 18018
In such conditions, you must first identify the bottleneck, and then take action to mitigate the congestion.
Identifying the congestion bottleneck
To identify the bottleneck in the Brocade VDX network, enter the show interface command at various
locations, and identify interfaces with incrementing TX and RX discards. Depending upon the TX or RX
discards, the congestion could be anywhere downstream.
Mitigating the congestion
Try the following actions to mitigate congestion:
• Increase bottleneck bandwidth.
‐ Add more links to the LAG and ECMP paths.
‐ Use higher-speed interfaces.
• Implement flow control on the bottleneck and on neighboring devices.
• Implement QoS congestion management schemes.
‐ Classify, mark, and prioritize critical traffic.
‐ Modify scheduling schemes. Consider and compare the effects of using strict priority or
deficit weighted round-robin (DWRR) scheduling schemes.
NIC teaming with vLAG
Network OS Administrator’s Guide 675
53-1003225-04