Cisco WS-SUP32-GE-3B - Supervisor Engine 32 Software Configuration Manual page 554

Software configuration guide
Hide thumbs Also See for WS-SUP32-GE-3B - Supervisor Engine 32:
Table of Contents

Advertisement

Understanding How PFC QoS Works
Ingress and Egress Buffers and Layer 2 CoS-Based Queues
The Ethernet LAN module port ASICs have buffers that are divided into a fixed number of queues. When
congestion avoidance
queues. The buffers and queues store frames temporarily as they transit the switch. PFC QoS allocates
the port ASIC memory as buffers for each queue on each port.
The Catalyst 6500 series switch LAN modules support the following types of queues:
The Catalyst 6500 series switch LAN modules support the following types of scheduling algorithms
between queues:
The Catalyst 6500 series switch LAN modules provides congestion avoidance with these types of
thresholds within a queue:
Catalyst Supervisor Engine 32 PISA Cisco IOS Software Configuration Guide, Release 12.2ZY
38-20
is enabled, PFC QoS uses the traffic's Layer 2 CoS value to assign traffic to the
Standard queues
Strict-priority queues
Shaped round robin (SRR)—SRR allows a queue to use only the allocated bandwidth.
Deficit weighted round robin (DWRR)—DWRR keeps track of any lower-priority queue
under-transmission caused by traffic in a higher-priority queue and compensates in the next round.
Weighted Round Robin (WRR)—WRR does not explicitly reserve bandwidth for the queues. Instead,
the amount of bandwidth assigned to each queue is user configurable. The percentage or weight allocated
to a queue defines the amount of bandwidth allocated to the queue.
Strict-priority queueing—Strict priority queueing allows delay-sensitive data such as voice to be
dequeued and sent before packets in other queues are dequeued, giving delay-sensitive data preferential
treatment over other traffic. The switch services traffic in the strict-priority transmit queue before
servicing the standard queues. After transmitting a packet from a standard queue, the switch checks
for traffic in the strict-priority queue. If the switch detects traffic in the strict-priority queue, it
suspends its service of the standard queue and completes service of all traffic in the strict-priority
queue before returning to the standard queue.
Weighted Random Early Detection (WRED)—On ports with WRED drop thresholds, frames with a
given QoS label are admitted to the queue based on a random probability designed to avoid buffer
congestion. The probability of a frame with a given QoS label being admitted to the queue or
discarded depends on the weight and threshold assigned to that QoS label.
For example, if CoS 2 is assigned to queue 1, threshold 2, and the threshold 2 levels are 40 percent
(low) and 80 percent (high), then frames with CoS 2 will not be dropped until queue 1 is at least
40 percent full. As the queue depth approaches 80 percent, frames with CoS 2 have an increasingly
higher probability of being discarded rather than being admitted to the queue. Once the queue is over
80 percent full, all CoS 2 frames are dropped until the queue is less than 80 percent full. The frames
the switch discards when the queue level is between the low and high thresholds are picked out at
random, rather than on a per-flow basis or in a FIFO manner. This method works well with protocols
such as TCP that can adjust to periodic packet drops by backing off and adjusting their transmission
window size.
Tail-drop thresholds—On ports with tail-drop thresholds, frames with a given QoS label are
admitted to the queue until the drop threshold associated with that QoS label is exceeded;
subsequent frames of that QoS label are discarded until the threshold is no longer exceeded. For
example, if CoS 1 is assigned to queue 1, threshold 2, and the threshold 2 watermark is 60 percent,
then frames with CoS 1 will not be dropped until queue 1 is 60 percent full. All subsequent CoS 1
frames will be dropped until the queue is less than 60 percent full. With some port types, you can
configure the standard receive queue to use both a tail-drop and a WRED-drop threshold by mapping
a CoS value to the queue or to the queue and a threshold. The switch uses the tail-drop threshold for
Chapter 38
Configuring PFC QoS
OL-11439-03

Hide quick links:

Advertisement

Table of Contents
loading

This manual is also suitable for:

Catalyst supervisor engine 32 pisa

Table of Contents