RPM-XF Installation and Configuration Guide, Release 3

Configuring Quality of Service

This chapter explains how to configure Quality of Service (QoS) on the RPM-XF, which supports the following features:

•Committed access rate (CAR) measures traffic rates and, based on the rates, takes actions (such as dropping packets). RPM-XF QoS supports CAR ("police") on input packets and shaping ("shape") on output packets.

•Random Early Detection (RED) is a congestion avoidance mechanism that takes advantage of TCP's congestion control mechanism. By randomly dropping packets prior to periods of high congestion, RED tells the packet source to decrease its transmission rate. Assuming the packet source is using TCP, it will decrease its transmission rate until all the packets reach their destination, indicating that the congestion is cleared.

•Weighted random early detection (WRED) uses an algorithm to randomly discard packets during congestion. This approach reduces congestion by causing the packet source to slow down.

Weighted RED (WRED) generally drops packets selectively based on IP precedence. Packets with a higher IP precedence are less likely to be dropped than packets with a lower precedence. Thus, higher priority traffic is delivered with a higher probability than lower priority traffic.

•Bandwidth reservation, also referred as fair queueing, assigns bandwidth to certain streams of packets.

•Low-latency priority queueing can be assigned for real-time traffic such as voice and video.

•Traffic shaping is used to control traffic by maintaining data flow at a set rate.

•Set specifies an IP precedence/DSCP or MPLS experimental value that can be used by other routers to manage QoS.

•802.1q support allows PXF switching for ARPA encapsulation.

In addition, the RPM-XF supports QoS policy propagation through the Border Gateway Protocol (QPPB). For a QPPB configuration example, see "Quality of Service Policy Propagation Example Using Border Gateway Protocol" section

General QoS Configuration Procedure

You can configure WRED, CAR, and other qualities of service by performing the following tasks:

1. Create a QoS boilerplate that defines the criteria for prioritizing traffic.

2. Apply the boilerplate to an interface.

Figure 10-1 shows an overview of the QoS process.

Figure 10-1 QoS Process

Creating a QoS Boilerplate

This section provides the information you need to create a QoS boilerplate. To create a QoS boilerplate, perform two procedures:

1. Create a class map—The class map tells the RPM-XF how to recognize the packets that are subject to QoS.

2. Create a policy map—The policy map lists QoS services to be applied to packets described by one or more class maps.

Creating a Class Map

The following procedure describes how to create a class map.

Step 1 Assign a name to your class map by entering the class-map name command. In the following example, a class map named mink is created.

Router(config)# class-map mink

Router(config-cmap)#

As the example shows, after you enter the class-map name command, you enter class map configuration mode (config-cmap).

Note Some Cisco IOS documents refer to the QoS configuration modes as the modular CLI.

Step 2 Describe the characteristics of the packets that are subject to QoS by entering the match command. In the following example, the packet is described as being associated to access group 10 and having the IP precedence bit set to 1.

Router(config-cmap)# match access-group 10 

Router(config-cmap)# match ip precedence 1

Step 3 Exit class map configuration mode.

Router(config-cmap)# exit

Router(config)

As a result of the creation of a class map, the router can recognize packets that are subject to QoS. You must now tell the router the action to take on those packets.

Creating a Policy Map

The following procedure describes how to create a policy map.

Step 1 Assign a name to your policy map by entering the policy-map name command. In the following example, a policy map named lynx is created.

Router(config)# policy-map lynx

Router(config-pmap)#

As the example shows, after you enter the policy-map name command, you enter policy map configuration mode (config-pmap).

Step 2 Associate the policy map with a class map.

Router(config-pmap)# class mink

Router(config-pmap-c)#

As the example shows, after entering the class name command, enter the policy map class configuration mode (config-pmap-c).

Step 3 Describe the QoS actions you want the router to perform when the router encounters a packet that has the characteristics described by the class map.

In this example, the router executes default behavior for the police command. (See the "Specifying a Committed Access Rate" section for details.)

Router(config-pmap-c)# police 80000

Step 4 Exit policy map configuration mode.

Router(config-pmap-c)# exit

Router(config-pmap)# exit

Router(config)#

You have completed the creation of a QoS boilerplate, which can be assigned to an interface.

Assigning a QoS Boilerplate to an Interface

Use the service-policy command to assign a QoS boilerplate to an interface. In the following example, the policy map lynx is assigned to traffic that enters the gigabit Ethernet interface of an RPM-XF.

Router(config)# interface gigabitethernet 1/0

Router(config-if)# service-policy input lynx

Class Map Commands

This section describes commands for creating and modifying class maps.

Creating a Class Map

You can create a class map and enter class-map configuration mode by entering the class-map command.

class-map [match-any | match-all] class-map-name

[no] class-map class-map-name

The default value is match-all.

Use the no class-map command to delete a class map.

Cisco IOS software supports a maximum of 255 unique class maps.

In the following example, a class-map named mink is created. In the example, the default value of match-all is used.

Router(config)# class-map mink

Router(config-cmap)# 

Matching Attributes

Use the match command to define the characteristics of the packets that belong to the class map.

match match_statement

[no] match match_statement

The match command match_statement is one of the following values:

•match [not] access-group number—Specifies that the packet must (or must not) be permitted by the access group whose number is from 1 to 2699.

•match [not] access-group name access-list-name—Specifies that the packet must (or must not) be permitted by the access list whose name is access-list-name.

•match [not] any—Specifies that all (or no) packets belong to this class.

•match [not] ip dscp code-point-value1 [º[code-point_value8]]—Specifies that the packet IP differentiated service code point (dscp) value must (or must not) match one or more of the code-point values in the range 0 to 63. You can specify up to eight code point values, separating consecutive values with a space.

•match [not] ip precedence prec_value1 [º[prec_value8]]—Specifies that the packet IP precedence value must (or must not) match one or more precedence values in the range 0 to 7. You can specify up to eight precedence values, separating consecutive values with a space.

•match [not] qos-group number—Specifies that the packet QoS group number value must (or must not) be in the group number range 0 to 99.

•match [not] ip rtp lowest-udp-port:2000-65535 range:0-16383—Used to match a packet that has an even numbered UDP port within (or outside of) the specified range. Only even ports are matched because they carry the real time data streams. Odd ports are not matched because they only carry control information.

•match [not] mpls experimental experimental-value—Specifies that the packet experimental bits must (or must not) match the specified experimental value(s) (in the range from 0 to 7). You can specify up to eight experimental values, separating consecutive values with a space. Note that this match can only match MPLS (tagged) frames.

Use the no form of this command to disable the match attributes.

To construct your mapping rules, enter one or more match commands. Each packet is compared to the criteria specified by the match commands to determine if the packet contains the attribute you specify.

The RPM-XF supports a maximum of 16 match statements in a class map.

In the following example, a class map is created that tells the router to look for packets that belong to access group 1 and have an IP precedence value of 3 or 7.

Router(config)# class-map mink

Router(config-cmap)# match access-group 1

Router(config-cmap)# match ip precedence 3 7

Policy Map Commands

This section describes commands for creating and modifying policy maps.

Creating a Policy Map

You can create a policy map and enter policy-map configuration mode by entering the policy-map command from global configuration mode.

policy-map policy-map-name

[no] policy-map policy-map-name

The policy-map-name can be any word or number.

Use the no form of the command to remove a policy map.

In the following example, a policy map named lynx is created.

Router(config)# policy-map lynx

Router(config-pmap)#

Assigning a Class to a Policy Map

Use the class class-map-name command from policy-map configuration mode to assign a class map to a policy map.

class class-map-name

[no] class class-map-name

The class-map-name is the name assigned to the class map.

Use the no form of the command to remove a class.

You can use a special class map name called class-default on a given interface to assign QoS policies to all packets that are not already described in the policy map by a class of a different name.

After you enter the class class-map-name command, you enter policy-map class configuration mode, in which you can enter QoS policies.

Tip A packet is processed by a policy map as soon as a match is found. When you assign class names to a policy map, assign the first name to the class that is most likely to be used. This can improve QoS performance.

In the following example, the class map named mink is assigned to the policy map named lynx.

Router(config)# policy-map lynx

Router(config-pmap)# class mink

Router(config-pmap-c)#

In the following example, the default class map is assigned to the policy map named lynx.

Router(config)# policy-map lynx

Router(config-pmap)# class class-default

Router(config-pmap-c)#

Specifying a Committed Access Rate

To specify a committed access rate (CAR), enter the police command while you are in policy-map class configuration mode. You can use this command to control low-priority traffic, so that an interface has more bandwidth for high-priority traffic or to enforce a specific rate on an interface.

police bps [normal-burst | max-burst | conform-action action | exceed-action action]

[no] police [normal-burst | max-burst | conform-action action | exceed-action action]

If you enter only police bps at the command line, the following default behavior occurs: traffic that conforms to the bps value is transmitted and traffic that exceeds the bps value is dropped.

Use the no form of the command to disable policing.

In the following example, CAR is assigned to the class named mink.

Router(config)# policy-map lynx

Router(config-pmap)# class mink

Router(config-pmap-c)# police 720000 90000 90000 conform-action transmit exceed-action 
drop

Enabling Weighted Random Early Detection

Use the random-detect command to enable weighted random early detection (WRED), which randomly discards packets during congestion based on IP precedence settings. The random-detect command enables a WRED drop policy for a traffic class that includes a bandwidth guarantee.

Note The bandwidth must be set before you can enable WRED (see Bandwidth Reservation and Low-Latency Priority Queueing).

Note On the ATM interface, you can only use WRED on a variable bit rate (VBR) PVCs. You cannot use WRED on PVCs configured for an unspecified bit rate (UBR).

random-detect [ewc value | prec prec-value min-value max-value mark-denom]

[no] random-detect [ewc value | prec prec-value min-value max-value mark-denom]

Tip In most cases, the benefits of WRED can be best realized if you use the random-detect keyword without arguments.

Use the no form of the command to disable WRED.

The following example shows the implementation of WRED.

Router(config)# policy-map lynx

Router(config-pmap)# class class-default

Router(config-pmap-c)# random-detect

Bandwidth Reservation and Low-Latency Priority Queueing

This section explains how to configure bandwidth reservation and low-latency priority. These queueing methods let you offer differentiated service to customers.

The RPM-XF typically uses a single queue for packets from all traffic streams waiting for the link to transmit them in the order of their arrival. This method is simple, efficient, and offers optimal average delay per packet because it always uses the entire link bandwidth. But the single queue method does not distinguish among different traffic streams—the more traffic in a stream, the larger its share of the link bandwidth.

Bandwidth reservation divides the link bandwidth among the different traffic streams into multiple queues, with each queue receiving its fair share of the link bandwidth divided among all non-empty queues. You do not waste bandwidth associated with an empty queue, and by dividing the unused bandwidth to the queues with packets to send, multiple queueing has the same average delay per packet as the single queue scheme, with the advantage of fairness.

Low-latency priority queueing lets you assign a guaranteed minimum bandwidth to one queue to minimize the packet-delay variance for delay-sensitive traffic, such as live voice and video.

Note Bandwidth and low-latency priority cannot be combined in the same class.

Bandwidth Reservation Queueing

Use the bandwidth command to create multiple class queues.

bandwidth rate-in-kbps

[no] bandwidth

The rate-in-kbps parameter is a value in the range from 8 to 2,000,000 representing between 1% to 99% of the link bandwidth.

Use the no form of the command to disable bandwidth queueing.

The following sample configuration creates two class queues.

•A 18 kbps queue for packets with IP precedence bit settings of 1, 2, 3, or 4.

•A 54 kbps queue for packets with IP precedence bit settings of 5, 6, or 7.

Assuming that the interface has 128 kbps bandwidth, the two class queues receive 25% and 62% of the interface bandwidth. All other traffic, including IP precedence 0, receives the rest of the bandwidth—8 kbps or 13%.

Router# enable

Router# configure terminal

Router(config)# class-map city

Router(config-cmap)# match ip precedence 1 2 3 4

Router(config-cmap)# class-map boston

Router(config-cmap)# match ip precedence 5 6 7

Router(config-cmap)# policy-map precedence-queues

Router(config-pmap)# class city

Router(config-pmap-c)# bandwidth 16

Router(config-pmap-c)# class boston

Router(config-pmap-c)# bandwidth 40

Router(config-pmap-c)# interface switch1:1

Router(config-if)# service-policy output precedence-queues

Router(config-if)# end

The actual throughput of a queue may be higher when one or more of the other queues on the link are idle.

Low-Latency Priority Queueing

Low-latency priority queueing lets you assign a specified share of the link bandwidth to one queue that receives priority over all others. Low-latency priority queueing minimizes the packet-delay variance for delay-sensitive traffic, such as live voice and video.

Use the priority command to create a low-latency priority queue.

priority rate-in-kbps

[no] priority

The rate-in-kbps parameter is a value in the range from 8 to 2,000,000, representing the guaranteed minimum bandwidth.

Use the no form of the command to disable priority queueing.

The following sample configuration creates a priority queue for voice traffic, and applies it to interface switch1.1.

Router# enable

Router# configure terminal

Router(config)# class-map voice

Router(config-cmap)# match ip rtp 2000 2000

Router(config-cmap)# policy-map voice-queue

Router(config-pmap)# class voice

Router(config-pmap-c)# priority 56

Router(config-pmap-c)# interface switch1:1

Router(config-if)# service-policy output voice-queue

Router(config-if)# end

The actual throughput of a priority queue may be higher than the minimum because it allocates the entire link bandwidth to a priority queue if all the other queues on the link are empty.

Generic Traffic Shaping

The RPM-XF uses traffic shaping as a mechanism to control or modify the flow of traffic on an interface to meet the requirements of a remote site, or to conform to a service rate that is provided on that interface. Generic Traffic Shaping (GTS) supports traffic shaping on all interfaces regardless of the encapsulation of the interface.

There are two implementations of traffic shaping in the current Cisco IOS software: GTS and Frame Relay Traffic Shaping (FRTS). This section describes GTS.

Note The RPM-XF does not support Frame Relay.

Note Use the traffic shape command in policy map class configuration mode. It is not supported in interface configuration mode.

The traffic shape command limits the throughput equal to rate-in-kbps.

shape rate-in-kbps

[no] shape

The rate-in-kbps parameter is a value in the range from 56 to 2,000,000, representing the maximum throughput allowed.

Use the no form of the command to disable traffic shaping.

In the following sample configuration, the traffic shape is set to a throughput of 100.

Router(config-pmap-c)# shape 100

Router(config-pmap-c)#

Specifying a Queue Limit

This section describes how to specify the number of packets held by the queue. Increase the queue limit to reduce the number of packets dropped due to temporary congestion on the assigned interface. Queue limit operates on the default packet drop method of congestion management.You cannot use the queue limit command on ATM PVCs configured for unspecified bit rate (UBR).

Note On the ATM interface, you can only apply queue limits on variable bit rate (VBR) PVCs.

queue-limit packets

[no] queue-limit

The packets parameter is a number of packets from 32 to 16,384 in powers of 2 (for example, 64, 128, 256).

Note If the number of packets specified is not a power of 2, the number entered is automatically rounded up to a power of 2. For example, if the number of packets is entered as 60, it will be rounded up to 64.

Use the no form of the command to return the queue limit to its default value.

Use the show interface command to determine the current queue limit. If you set the queue limit to a high value, this may reduce the number of packet buffers available to other interfaces.

In the following example, the queue limit is set to 256 packets:

Router(config)# policy-map lynx

Router(config-pmap)# class class-default

Router(config-pmap-c)# queue-limit 256

Applying Set Values

The set command allows you to mark bit values that can be used by other routers to manage QoS.

set {ip {dscp value | precedence value} | qos-group value | atm-clp | mpls experimental 
value}

[no] set {ip {dscp value | precedence value} | qos-group value | mpls experimental value}

Externally visible values are the following.

Internally visible values are the following.

Use the no form of the command to return the set values to their defaults.

In the following sample configuration, bit values are set for IP precedence and a QoS group.

Router(config)# policy-map lynx

Router(config-pmap)# class mink

Router(config-pmap-c)# set ip precedence 7

Router(config-pmap-c)# set qos-group 8

Service-Policy Command

To associate a policy map with an interface, use the service-policy command.

service-policy [input | output] name

[no] service-policy [input | output] name

Note The bandwidth, low-latency priority, random-delete, queue limit, and shape parameters are used with output only, and are ignored when using the input argument.

Use the no form of the command to remove a service policy from an interface.

No more than two service policies can be associated with an interface, one for input and one for output.

On the ATM interface, you can only apply a policy map on a PVC.

In the following example, the policy map lynx is applied to the incoming traffic on an interface of the Gigabit Ethernet line card.

Note CEF switching must be on to use the service-policy command.

Router(config)# interface gigabitethernet 1/0

Router(config-if)# service-policy input lynx

In the following example, a policy map is applied to an ATM PVC.

Router(config)# interface switch1.1

Router(config-if)# pvc 0/101

Router(config-if-atm-vc)# service-policy input lynx

Show Commands

This section lists show commands you can use to get information about class maps and policy maps.

show policy map

This command displays the configuration of one or all policy maps and lists information about the configurations. For example,

Router# show policy-map lynx

 Policy Map lynx

  class  mink

   set qos-group 8

 Policy Map jaguar

  class  class-default

   random-detect

   random-detect exponential-weighting-constant 9

   random-detect precedence 0 16 32 10

   random-detect precedence 1 18 32 10

   random-detect precedence 2 20 32 10

   random-detect precedence 3 22 32 10

   random-detect precedence 4 24 32 10

   random-detect precedence 5 26 32 10

   random-detect precedence 6 28 32 10

   random-detect precedence 7 30 32 10

show policy-map interface

This command displays statistics of a policy map on one or all interfaces. This example shows statistics for a particular serial interface.

Router# show policy-map interface [pos1/0]

  Pos1/0 

  service-policy input: lynx

    class-map: mink (match-all)

      0 packets, 0 bytes

      5 minute rate 0 bps

      match: access-group 3

      set:

        qos-group 8

show class-map

This command lists the class maps and displays their match statements. For example,

Router# show class-map mink

 Class Map match-all mink (id 3)

   Match access-group  3

 Class Map match-all pink (id 4)

   Match access-group  23

   Match qos-group  32

 Class Map match-any class-default (id 0)

   Match any

 Class Map match-all customer_pri (id 2)

show vlans

This command can list up to 1000 virtual LAN subinterfaces. For example,

Router# show vlans

Virtual LAN ID: 1 (IEEE 802.1Q Encapsulation)

VLAN Trunk Interface:  GigabitEthernet1/0

Protocols Configured:  Address:       Received:  Transmitted:

       IP              200.1.1.1      18         273894058

Quality of Service Policy Propagation Example Using Border Gateway Protocol

Quality of Service (QoS) Policy Propagation using Border Gateway Protocol (QPPB) allows you to classify packets by IP precedence based on BGP community lists, BGP autonomous system paths, and access lists. After a packet has been classified, you can use other QoS features such as committed access rate (CAR) and weighted random early detection (WRED) to specify and enforce policies to fit your business model.

The example below shows how t odo the following.

1. Create route maps to match BGP community lists, access lists, and BGP AS paths.

2. Apply IP precedence to routes learned from neighbors.

In this example, the RPM-XF learns routes from autonomous system (AS) 10 and AS 60. QoS policy is applied to all packets that match the defined route maps. Any packets from the RPM-XF to AS 10 or AS 60 are sent to the appropriate QoS policy (see Figure 10-2).

Figure 10-2 RPM-XF Routes and QoS Policy Application

RPM-XF Configuration

Router(config)# router bgp 30

Router(config)# table-map precedence-map

Router(config-router)# neighbor 20.20.20.1 remote-as 10

Router(config-router)# neighbor 20.20.20.1 send-community

Router(config-router)# neigh 20.20.20.1 route-map precedence-map out

!

Router(config)# ip bgp-community new-format

Match community 1, set the IP precedence to priority, and set the QoS group to 1.

Router(config)# route-map precedence-map permit 10

Router(config-route-ma)# match community 1

Router(config-route-ma)# set ip precedence priority

Router(config-route-ma)# set ip qos-group 1

Match community 2 and set the IP precedence to immediate.

Router(config)# route-map precedence-map permit 20

Router(config-route-ma)# match community 2

Router(config-route-ma)# set ip precedence immediate

Match community 3 and set the IP precedence to Flash.

Router(config)# route-map precedence-map permit 30

Router(config-route-ma)# match community 3

Router(config-route-ma)# set ip precedence flash

Match community 4 and set the IP precedence to Flash-override.

Router(config)# route-map precedence-map permit 40

Router(config-route-ma)# match community 4

Router(config-route-ma)# set ip precedence flash-override

Match community 5 and set the IP precedence to critical.

Router(config)# route-map precedence-map permit 50

Router(config-route-ma)# match community 5

Router(config-route-ma)# set ip precedence critical

Match community 6 and set the IP precedence to internet.

Router(config)# route-map precedence-map permit 60

Router(config-route-ma)# match community 6

Router(config-route-ma)# set ip precedence internet

Match community 7 and set the IP precedence to network.

Router(config)# route-map precedence-map permit 70

Router(config-route-ma)# match community 7

Router(config-route-ma)# set ip precedence network

Match ip address access list 69 or match AS path 1, set the IP precedence to critical, and set the QoS group to 9.

Router(config)# route-map precedence-map permit 75

Router(config-route-ma)# match ip address 69

Router(config-route-ma)# match as-path 1

Router(config-route-ma)# set ip precedence critical

Router(config-route-ma)# set ip qos-group 9

For everything else, set the IP precedence to routine.

Router(config)# route-map precedence-map permit 80

Router(config-route-ma)# set ip precedence routine

Define the community lists.

Router(config)# ip community-list 1 permit 60:1

Router(config)# ip community-list 2 permit 60:2

Router(config)# ip community-list 3 permit 60:3

Router(config)# ip community-list 4 permit 60:4

Router(config)# ip community-list 5 permit 60:5

Router(config)# ip community-list 6 permit 60:6

Router(config)# ip community-list 7 permit 60:7

Define the AS path.

Router(config)# ip as-path access-list 1 permit ^10_60

Define the access list.

Router(config)# access-list 69 permit 69.0.0.0

Router B Running Configuration

RouterB(config)# router bgp 10

RouterB(config-router)# neighbor 30.30.30.1 remote-as 30

RouterB(config-router)# neighbor 30.30.30.1 send-community

RouterB(config-router)# neigh 30.30.30.1 route-map send_community out

!

RouterB(config)# ip bgp-community new-format

Match prefix 10 and set community to 60:1.

RouterB(config)# route-map send_community permit 10

RouterB(config-route-ma)# match ip address 10

RouterB(config-route-ma)# set community 60:1

Match prefix 20 and set community to 60:2.

RouterB(config)# route-map send_community permit 20

RouterB(config-route-ma)# match ip address 20

RouterB(config-route-ma)# set community 60:2

Match prefix 30 and set community to 60:3.

RouterB(config)# route-map send_community permit 30

RouterB(config-route-ma)# match ip address 30

RouterB(config-route-ma)# set community 60:3

Match prefix 40 and set community to 60:4.

RouterB(config)# route-map send_community permit 40

RouterB(config-route-ma)# match ip address 40

RouterB(config-route-ma)# set community 60:4

Match prefix 50 and set community to 60:5.

RouterB(config)# route-map send_community permit 50

RouterB(config-route-ma)# match ip address 50

RouterB(config-route-ma)# set community 60:5

Match prefix 60 and set community to 60:6.

RouterB(config)# route-map send_community permit 60

RouterB(config-route-ma)# match ip address 60

RouterB(config-route-ma)# set community 60:6

Match prefix 70 and set community to 60:7.

RouterB(config)# route-map send_community permit 70

RouterB(config-route-ma)# match ip address 70

RouterB(config-route-ma)# set community 60:7

For all others, set community to 60:8.

RouterB(config)# route-map send_community permit 80

RouterB(config-route-ma)# set community 60:8

Define the access lists.

RouterB(config)# access-list 10 permit 61.0.0.0

RouterB(config)# access-list 20 permit 62.0.0.0

RouterB(config)# access-list 30 permit 63.0.0.0

RouterB(config)# access-list 40 permit 64.0.0.0

RouterB(config)# access-list 50 permit 65.0.0.0

RouterB(config)# access-list 60 permit 66.0.0.0

RouterB(config)# access-list 70 permit 67.0.0.0

The following example shows how to configure several interfaces to classify packets based on the IP precedence and QoS group ID.

interface switch1.1

ip address 200.28.38.2 255.255.255.0

bgp-policy source ip-prec-map

no ip mroute-cache

no cdp enable

frame-relay interface-dlci 20 IETF

interface switch1.2

ip address 200.28.28.2 255.255.255.0

bgp-policy source qos-group

no ip mroute-cache

no cdp enable