Configuring CoS Queues in Layer 2 Circuit Trunk Mode
On ATM2 IQ interfaces, you can configure ATM CoS scheduling for AAL5 mode and Layer 2 circuit trunk mode. For general information about ATM CoS, see ATM2 IQ VC Tunnel CoS Components Overview.
When you configure CoS scheduling in Layer 2 circuit trunk mode, the trunk is defined on the logical interface, and four CoS queues are opened in the trunk. For each CoS queue, you specify a priority and a transmit weight. CoS queues are serviced using a weighted round robin (WRR) algorithm. One queue is serviced with strictly high priority and the remaining queues are serviced with the WRR.
For Layer 2 circuit trunk mode, only strict mode is supported. Alternate mode is not supported.
To configure CoS queues in Layer 2 circuit trunk mode, perform the following tasks:
- Include the encapsulation atm-ccc-cell-relay statement at the [edit interfaces at-fpc/pic/port] hierarchy
level:[edit interfaces at-fpc/pic/port]encapsulation (atm-ccc-cell-relay | ether-vpls-over-atm-llc);
- Include the scheduler-maps statement at the [edit interfaces at-fpc/pic/port atm-options] hierarchy level:
- Include the atm-scheduler-map, trunk-bandwidth, and trunk-id statements at the [edit interfaces
at-fpc/pic/port unit logical-unit-number] hierarchy level:[edit interfaces at-fpc/pic/port unit logical-unit-number]atm-scheduler-map (map-name | default);trunk-bandwidth rate;trunk-id number;
For information about ATM scheduler maps, see Configuring an ATM Scheduler Map.
For information about trunk identification numbers, see Configuring Layer 2 Circuit Transport Mode. For information about trunk bandwidths, see Configuring Layer 2 Circuit Trunk Mode Scheduling Overview.
Strict mode CoS queue priority works as follows:
Scheduling—One queue has strictly high priority and is always serviced before the remaining queues are serviced by a weighted round robin. This means the packets in a high priority queue are sent first until the queue is empty. Then low priority queues send packets until their weight quota becomes zero or negative.
Latency—Each trunk is associated with a trunk bandwidth. The trunk bandwidth is the maximum bandwidth used each time a trunk is serviced. In the scheduling process, each trunk is serviced in a WRR. The maximum latency for any trunk to begin transmitting is equal to the sum of the weights of all previously queued trunks. Trunks without data do not affect output scheduling. As long as all the trunks have data, the exact weight proportions are maintained. If a trunk runs out of data during its turn, it is no longer included in the WRR. When the trunk gets more data, the trunk is placed at the end of the queue. For more information, see Configuring Layer 2 Circuit Trunk Mode Scheduling Overview.
Within a single trunk, the maximum latency of a high priority queue is the time it takes to transmit one ATM cell. The latency of a low priority queue is the sum of high priority queue burst time and the transmission time of the remaining low priority queues’ weight.
Bandwidth distribution—Trunks are serviced in a WRR based on the trunk bandwidth.
Within a single trunk, the high priority queue consumes the bandwidth first regardless of its weight. The remaining bandwidth is distributed to the low priority queues in proportion to their weights.
Consider the following example:
You configure a trunk with weights of 10 percent, 20 percent, 30 percent, and 40 percent for queues 0, 1, 2, and 3, respectively.
You configure queue 0 to be a high priority queue.
Queue 0 does not have cells to transmit.
In this scenario, queues 1, 2 and 3 receive 2/9, 3/9, and 4/9 of the bandwidth, respectively.
Constant bit rate (CBR) traffic always enters the strictly high priority queue.
For more information about strict and alternate modes, see ATM2 IQ VC Tunnel CoS Components Overview.
For general information about Layer 2 circuit trunk mode, see Configuring Layer 2 Circuit Transport Mode.
For interfaces configured in trunk mode, you can also configure dual EPD thresholds depending on packet loss priorities (PLPs). For more information, see Configuring Two EPD Thresholds per Queue.