Congestion Management and Avoidance (1/3)

Search in book...
Toggle Font Controls
Create new playlist

Name your new playlist

Playlist description (optional)
Sign In

Email address

Password

Forgot Password?

or

Continue with Facebook

Continue with Google
Sign Up

Full Name

Email address

Confirm Email Address

Password

or

Continue with Facebook

Continue with Google

Congestion Management and Avoidance 175

Congestion Management and Avoidance

After the switch classiﬁes, marks, and polices a packet against QoS rules, the switch queues

the packet into a transmit queue for output scheduling. Placing the packet into one of

multiple queues allows for the switch to differentiate service by transmitting packets from

the queue with speciﬁc order and priority. The Catalyst 3550 Family of switches utilizes

WRED and tail-drop queuing mechanisms for congestion avoidance; the Catalyst 2950

Family of switches does not support congestion avoidance. Both the Catalyst 2950 Family

and 3550 Family of switches support four transmit queues. Although both switches support

output scheduling utilizing WRR, the Catalyst 3550 Family of switches supports additional

output scheduling and queuing conﬁgurations.

In brief, the Catalyst 2950 Family of switches supports the following output scheduling

mechanism:

• Strict-priority scheduling

• WRR scheduling

Strict-priority scheduling services packets placed into higher-priority queues before

servicing packets in lower-priority queues. Although this mechanism works well for high-

priority trafﬁc such as Voice over IP (VoIP), this mechanism may starve transmission of

trafﬁc in lower-priority queues. As a result, the second option of using WRR scheduling

exists. WRR transmits trafﬁc based on a weight value. In this manner, each queue receives

an assigned weight of the total bandwidth. Therefore, during any period of time, the switch

sends trafﬁc out of every queue based on its weighted value. Subsequent sections discuss

these mechanisms in more detail with conﬁguration examples.

In brief, the Catalyst 3550 Family of switches supports the following queuing and output

scheduling mechanisms:

• Expedite (strict-priority) queue

• WRR scheduling

• Conﬁgurable drop thresholds per output queue

• WRED congestion avoidance algorithm

The Catalyst 3550 Family of switches uses WRR scheduling for output scheduling in a

manner similar to the Catalyst 2950 Family of switches. However, the Catalyst 3550 Family

of switches allows for designation of an expedite queue and conﬁguration of congestion

avoidance mechanisms using tail-drop thresholds or WRED. The designation of the

expedite queue forces strict priority of transmission on the respective egress queue. The

congestion avoidance algorithms attempt to subside congestion by dropping packets with

lower priority before higher-priority packets in the same transmit queue. Later sections of

this chapter discuss these features in more detail.

176 Chapter 6: QoS Features Available on the Catalyst 2950 and 3550 Family of Switches

Congestion Management with the Catalyst 2950 and 3550 Family of

Switches

As noted in the preceding section, both the Catalyst 2950 Family and Catalyst 3550 Family

of switches perform output scheduling via strict-priority queuing and WRR, although each

family of switches exploits different default conﬁgurations. In summary, both the Catalyst

2950 Family and Catalyst 3550 Family of switches support the following congestion

management conﬁguration options:

• DSCP-to-CoS Mapping

• CoS-to-Transmit Queue Mapping

• Strict-Priority Queuing

• Conﬁgurable Weights for WRR

DSCP-to-CoS Mapping

The Catalyst 2950 Family and 3500 Family of switches use an internal DSCP value to

differentiate service as a packet traverses the switch. Because marking may occur on a

packet during the marking and policing of packet process, the CoS value of the packet needs

to be updated on transmit. As a result, the switch employs the use of a DSCP-to-CoS

mapping table for mapping the internal DSCP value to CoS value on packets before the

switch schedules the packet into a transmit queue. Table 6-5 indicates the default DSCP-to-

CoS mapping table.

The DSCP-to-CoS mapping table is conﬁgurable on a global basis. Use the following

global conﬁguration command to conﬁgure the DSCP-to-CoS mapping values:

mls qos map

dscp-cos dscp-list

cos

dscp-list represents up to eight DSCP values separated by a space. CoS corresponds to the

transmitted CoS value on the egress frame. Example 6-25 illustrates a user conﬁguring and

verifying the QoS DSCP-to-CoS mapping.

Table 6-5 Default DSCP-to-CoS Mapping Table

DSCP Value 0–7 8–15 16–23 24–31 32–39 40–47 48–55 56–63

CoS Value 01234567

Example 6-25 Verifying and Conﬁguring the DSCP-to-CoS Mapping

Switch#configure terminal

Switch(config)#mls qos map dscp-cos 37 38 39 to 5

Switch(config)#mls qos map dscp-cos 32 to 3

Switch(config)#end

Switch#show mls qos map dscp-cos

Dscp-cos map:

d1 : d2 0 1 2 3 4 5 6 7 8 9

Congestion Management and Avoidance 177

CoS-to-Transmit Queue Mapping

With the Catalyst 2950 and 3550 Family of switches, output scheduling uses the strict-

priority queuing or WRR algorithm with four transmit queues. The switch queues trafﬁc

into the four transmit queues based solely on the CoS value associated with the frame. By

default, the switch places frames with higher CoS values into the higher-number queues.

The internal DSCP value actually determines the CoS value and subsequently the transmit

queue. Before the switches places packets into the transmit queues, the switch determines

the egress CoS value of the frame using the internal DSCP value and the DSCP-to-CoS

mapping table.

In retracing the path of the packet, refer to the following outline and Figure 6-5 to under-

stand the role of internal DSCP in CoS-to-transmit queue mapping.

• Switch receives packet.

• Switch classiﬁes packets with an internal DSCP value based on conﬁguration.

• Switch polices, marks, or marks down the internal DSCP value.

• Switch determines egress CoS value based on internal DSCP value and DSCP-to-CoS

mapping table.

• Switch queues the packet into one of four transmit queues based on egress CoS value,

CoS-to-transmit queue mapping table, and any conﬁgured congestion avoidance

conﬁgurations.

• Switch schedules packet out of the transmit queue based on strict-priority queuing or

WRR.

Table 6-6 illustrates the default CoS-to-transmit queue mapping.

---------------------------------------

0 : 00 00 00 00 00 00 00 00 01 01

1 : 01 01 01 01 01 01 02 02 02 02

2 : 02 02 02 02 03 03 03 03 03 03

3 : 03 03 03 04 04 04 04 05 05 05

4 : 05 05 05 05 05 05 05 05 06 06

5 : 06 06 06 06 06 06 07 07 07 07

6 : 07 07 07 07

Table 6-6 Default CoS-to-Transmit Queue Mapping Table

CoS Value Transmit Queue

0, 1 1

2, 3 2

4, 5 3

6, 7 4

Example 6-25 Verifying and Conﬁguring the DSCP-to-CoS Mapping (Continued)

178 Chapter 6: QoS Features Available on the Catalyst 2950 and 3550 Family of Switches

Use the following command to display the CoS-to-transmit queue mapping:

The Catalyst 2950 Family of switches allows for custom conﬁguration of the CoS-to-

transmit queue mapping on a global basis. The Catalyst 3550 Family of switches supports

CoS-to-transmit queue mapping on a per-interface basis. Use the following global and

interface conﬁguration command to conﬁgure the CoS-to-transmit queue mapping:

wrr-queue cos-map

qid cos1..cosn

q_id represents one of the four transmit queues. cos1..cosn represents conﬁguration for up

to eight CoS values to map to a transmit queue.

Example 6-26 illustrates a user custom conﬁguring the CoS-to-transmit queue mapping

table and verifying the DSCP-TxQueue mapping table conﬁguration.

Strict-Priority Queuing

By default, the Catalyst 2950 Family of switches schedules packets from transmit queues

using the strict-priority algorithm. Using this algorithm, the switch transmits all packets out

of the higher-priority queues before servicing lower-priority queues.

The Catalyst 3550 Family of switches utilizes the strict-priority queue conﬁguration with

WRR for scheduling all the packets out of transmit queue four before servicing any other

queue. The switch still utilizes WRR for scheduling packets out of the remaining queues.

The behavior of this conﬁguration is slightly different from the behavior on the Catalyst

2950 Family of switches. Priority Queuing on the Catalyst 2950 Family of switches

services higher-priority queues before lower-priority queues in order and does not support

the concept of a single expedite queue.

To conﬁgure the Catalyst 3550 Family of switches for the strict-priority scheduling of

transmit queue four, use the following interface command:

As a result of the strict-priority scheduling algorithm on both the Catalyst 2950 and 3550

Family of switches, the switches may never transmit trafﬁc out of lower-priority queues

when line-rate trafﬁc exists in higher-priority queues. Although this behavior is warranted

for high-priority trafﬁc such as VoIP, many applications of strict-priority queuing may

starve lower-priority trafﬁc. To remedy this situation, consider conﬁguring the switch for

WRR bandwidth and WRED as discussed in the next section.

Example 6-26 User Conﬁguring and Verifying the DSCP-to-Transmit Queue Mapping Table

Switch#configure terminal

Switch(config)#wrr-queue cos-map 1 0 1 2

Switch(config)#wrr-queue cos-map 2 3 4

Switch(config)#wrr-queue cos-map 3 5

Switch(config)#end

Switch#show wrr-queue cos-map

CoS Value : 0 1 2 3 4 5 6 7

Priority Queue : 1 1 1 2 2 3 4 4

Congestion Management and Avoidance 179

Conﬁgurable Weights for WRR

By default, the Catalyst 3550 Family of switches utilizes WRR for congestion management.

The Catalyst 2950 Family of switches supports WRR as an optional conﬁguration. Both the

Catalyst 2950 Family and 3550 Family of switches support adjusting the scheduling weight

for each transmit queue. For the Catalyst 2950 Family of switches, WRR conﬁguration is a

global conﬁguration. On the Catalyst 3550 Family of switches, WRR conﬁguration is per

interface. The scheduling weight effectively determines the egress bandwidth per queue. Use the

following egress bandwidth formula to determine WRR weight:

(W/S )* B = n

W represents the WRR weight of the queue, and S represents the sum of all weights of the

active queues. B is the available bandwidth of the outgoing interface(s), and n represents

the egress bandwidth.

For example, each queue, 1 through 4, is assigned the following WRR weights on a Gigabit

Ethernet interface, respectively:

50 50 100 200

The S of the weights is 400. As a result of this conﬁguration, each transmit queue receives

the following egress bandwidth:

Queue 1: 125 Mbps

Queue 2: 125 Mbps

Queue 3: 250 Mbps

Queue 4: 500 Mbps

To conﬁgure the scheduling weight of total bandwidth per queue, use the following global

conﬁguration command for the Catalyst 2950 Family of switches and interface conﬁgu-

ration command for the Catalyst 3550 Family of switches:

wrr-queue bandwidth

weight1...weight4

weight1…weight4 represent four weighted values assigned to transmit queues 1, 2, 3, 4 respec-

tively. Use the following command to display the conﬁgured WRR bandwidth per queue:

Example 6-27 illustrates a user conﬁguring the WRR bandwidth per queue for heavier

weights on higher-priority queues and verifying the WRR bandwidth conﬁguration on a

Catalyst 2950 switch. In this example, queue 1 receives 10 percent of the bandwidth, queue 2

receives 20 percent of the bandwidth, queue 3 receives 30 percent of the bandwidth, and

queue 4 receives 40 percent of the bandwidth on egress.

Example 6-27 User Conﬁguring and Verifying a WRR Bandwidth Conﬁguration

Switch#configure terminal

Switch(config)#wrr-queue bandwidth 10 20 30 40

Switch(config)#end

Switch#show wrr-queue bandwidth

WRR Queue : 1 2 3 4

Bandwidth : 10 20 30 40

..................Content has been hidden....................

You can't read the all page of ebook, please click here login for view all page.

Table of Contents for Congestion Management and Avoidance (1/3)

Create new playlist

Sign In

Sign Up

Table of Contents for
Congestion Management and Avoidance (1/3)